understanding the complex aroma chemistry of …

UNDERSTANDING THE COMPLEX AROMA CHEMISTRY OF PREMIUM AGED RUMS

BY

CHELSEA MORGAN ICKES

DISSERTATION

Submitted in partial fulfillment of the requirements

for the degree of Doctor of Philosophy in Food Science and Human Nutrition

with a concentration in Food Science

in the Graduate College of the

University of Illinois at Urbana-Champaign, 2017

Urbana, Illinois

Doctoral Committee:

Professor Shelly J. Schmidt, Chair

Professor Keith R. Cadwallader, Director of Research

Professor Soo-Yeun Lee

Dr. Dawn M. Bohn

ii

Abstract

Rum is produced by the fermentation of sugar cane juice, syrup or molasses, followed by distillation

and then aging in oak barrels. Rum is a highly diverse distilled spirit because it has a somewhat

simple standard of identity, with the only requirement being that it must be produced from sugar

cane or its byproducts. The lack of regulation allows for manufacturers to pick and choose from a

variety of manufacturing practices when they are creating their rum. Rum cannot only be made from

different types of starting materials but variation exists in type of yeast and bacteria used for

fermentation, length of fermentation, distillation apparatus used, barrel type and length of

maturation. In today’s drink and bar culture, rum is experiencing a resurgence. High quality rums,

typically those aged at least five years and regarded as best of their class, are being compared with

fine spirits, such as Bourbon, Brandy, Cognac and Scotch. Therefore, the goal of the study was to

better understand the complex flavor chemistry of rum, including its aroma composition and the

effect of ethanol on flavor perception, with a main focus on premium aged rums. Nine rums were

evaluated consisting of two mixing rums (Bacardi Superior [BW], Bacardi Gold [BG]) and seven

premium rums (Appleton Estate V/X [AE], Appleton Estate Extra [AE12], Ron Abuelo: Añejo 7

years [RA7], Diplomatica Reserva Exclusiva [DR12], El Dorado 12 year old [ED12], Ron Zacapa

(Centenario) XO: Solera Gran Reserva Especial [RZ], Dictador XO Insolent [DX]).

Identification of the odor-active compounds in the nine rums by gas chromatography-olfactometry

(GCO) and GC-mass spectrometry (GC-MS) analysis yielded 59 odor-active regions containing 64

odor-active compounds. Aroma extract dilution analysis (AEDA) provided a ranking of the potency

of odorants. The most potent rum ordorants, although not necessasarily present in every rum, were

found to be acetal (melon), 2-/3-methyl-1-butanol (chocolate), β-damascenone (applesauce), 2-

phenethyl alcohol (roses), cis-whiskey lactone/4-methylguaiacol (sweet, coconut-like), eugenol (spicy,

clove), sotolon (curry, maple-like), syringol (smoky, spicy), (E)-isoeugenol (floral, clove), vanillin

(vanilla, sweet-like), ethyl vanillate (vanilla, sweet-like), and syringaldehyde (vanilla). Thirty-four of

the compounds identified by GCO and AEDA were quantitated by stable isotope dilution analysis.

Differences among the samples included the absence of 4-ethylguaiacol and eugenol in BW and the

presence of ethyl vanillin in only DR12 and DX. The mixing rums and DX were found to have the

lowest concentrations of all compounds quantitated in the rums. The quantitation results were

iii

converted to odor activity values (OAVs) to gain a better understanding of the importance of the

compounds to the overall aroma of the rum. Twenty-six compounds were found to have OAVs >1

in at least one rum. Fifteen compounds had OAVs >1 in all nine samples including 2-

methylpropanal, acetal, 3-methylbutanal, 2-methylbutanal, ethyl 2-methylpropanoate, ethyl

butanoate, ethyl 2-methylbutanoate, ethyl 3-methylbutanoate, 3-methyl-1-butanol, 2-methyl-1-

butanol, ethyl hexanoate, β-damascenone, guaiacol, cis-whiskey lactone and vanillin.

In order to characterize the sensory differences among rum products a rum flavor lexicon was

created through the use of web-based material. This is the first lexicon to be created for rum as well

as the first to use web-based materials for the lexicon development. The final lexicon consisted of

147 terms sorted into 22 categories. Descriptive sensory analysis was then conducted to verify the

rum flavor lexicon and to quantitate the sensory differences among nine rums previously evaluated

by analytical measures. Thirty-three of the 38 terms used to evaluate the rums were found on the

flavor wheel, validating that the lexicon contained terms relevant to the sensory evaluation of rums.

Twenty-three sensory attributes were found to be significantly different among rums. Two rums,

DX and DR12, were characterized by having higher intensity ratings for brown sugar, caramel,

vanilla and chocolate aroma, caramel, maple and vanilla aroma-by-mouth, and caramel aftertaste

compared to the other seven rums. Sensory profiles of the other seven rums were similar to one

another.

Descriptive analysis was also conducted to gain insight into the effect of ethanol on flavor

perception. Two rums, RA7 and DR12, were evaluated at three different dilution levels: straight

rum, 1:2 dilution with water, and a 1:2 dilution with 40%ABV. Dilutions of rum with water, while

hypothesized to alter the flavor profile of rum, yielded similar profiles to straight rum, except with

slightly lower attribute intensity ratings. However, dilution with 40% ethanol did significantly change

the profile of rum and also had the lowest intensity rating in the dilution series for most attributes.

Finally, chemometric analysis was conducted to correlate the sensory and analytical data using

principal component analysis consisting of quantitation, OAV and flavor dilution factor data.

Correlations between sensory evaluations with either quantitation or OAV data explained the most

variation among rums, accounting for 68.6% or 65.5%, respectively. Results indicate the changes in

vanilla, caramel, maple and chocolate aromas are driven by vanillin and ethyl vanillin. Additionally,

roasted aroma is defined by an absence of compounds rather than increases in concertation of any

iv

specific aroma compounds. Overall, the main differences between mixing and premium rums is the

concentrations of compounds, with mixing rums having lower concentrations of all compounds.

Differences in concentration and ratios of compounds relative to each other seem to be the driving

forces behind the difference in flavor perception among rums. These findings help to better

characterize rum as a category and articulate the differences that exist among rum categories.

Sensory evaluation of rum provides insight into how rums are perceived by the human senses and

the developed flavor lexicon will aid in communication between all levels of rum production and

consumers.

v

Acknowledgements

First and foremost, I would like to thank my advisor Dr. Keith Cadwallader, for his guidance,

encouragement, knowledge, and mentorship over the past five years. Without his support, none of

this would have been possible.

I would like to thank Dr. Shelly Schmidt, Dr. Soo-Yeun Lee and Dr. Dawn Bohn for their

willingness to serve on my advisory committee. Thank you for your wisdom, advice and

encouragement throughout this project.

I would also like to extend my gratitude and thanks to my labmates, past and present, for their

encouragement, assistance, and friendship, particularly Elizabeth Genthner, Samantha Gardiner,

Samantha McKenna, Yun Yin, and Bethany Hausch. Additionally, I would like to thank my peers

and friends in the Department of Food Science and Human Nutrition.

I would also like to thank my mentors from my internship at McCormick and Co. Inc, particularly

Paul Ford and Dana Gasiorowski, who introduced me to the world of flavor chemistry and that this

was something I could pursue as a career.

I would also like to thank my wonderful friends I have made outside of the university during my

time in here. You have not only encouraged and supported me but helped me grow as a person and

follower of Christ as well.

Finally, I would like to thank my family, especially my mom, dad and sisters, for your love and

support, not only through my graduate studies, but also throughout my entire life. Thank you

particularly to my mom, for moving across the country to watch James after he was born so I could

finish my degree. And last but not least, thank you to my husband Andrew and son James. Andrew,

thank you for your love, support, encouragement and inspiration to keep going and finish well.

Thank you for choosing to do life together. James, you are not only one of the best things to happen

to me during graduate school but my reason for finishing, to show you that you can do anything you

set your mind to.

vi

Table of Contents

List of Tables ..................................................................................................................................................... ix

List of Figures .................................................................................................................................................... xi

Chapter 1: Introduction .................................................................................................................................... 1

1.1 References ............................................................................................................................................... 5

Chapter 2: Literature Review ............................................................................................................................ 8

2.1 Introduction ............................................................................................................................................ 8

2.2 History and Production of Rum .......................................................................................................... 8

2.3 Rum Production ................................................................................................................................... 15

2.4 Flavor Research on Rum..................................................................................................................... 20

2.5 Sensory Studies on Rum ..................................................................................................................... 25

2.6 Flavor Wheels ....................................................................................................................................... 26

2.7 Effects of Ethanol on Flavor Perception of Alcoholic Beverages ............................................... 27

2.8 Figures ................................................................................................................................................... 32

2.9 References ............................................................................................................................................. 34

Chapter 3: Identification of Odor-Active Compounds in a Selection of Premium Rums ................... 43

3.1 Abstract ................................................................................................................................................. 43

3.2 Introduction .......................................................................................................................................... 43

3.3 Materials and Methods ........................................................................................................................ 45

3.4 Results and Discussion ........................................................................................................................ 48

3.5 Tables and Figure ................................................................................................................................. 57

3.6 References ............................................................................................................................................. 62

Chapter 4: Quantitation of Odor-Active Compounds in a Selection of Premium Rums and Use of

Chemometrics to Correlate Analytical and Sensory Analysis .................................................................... 68

4.1 Abstract ................................................................................................................................................. 68

4.2 Introduction .......................................................................................................................................... 68

4.3 Materials and Methods ........................................................................................................................ 70

4.4 Results and Discussion ........................................................................................................................ 78

4.5 Tables and Figures ............................................................................................................................... 89

4.6 References .......................................................................................................................................... 112

Chapter 5: Novel Creation of a Rum Flavor Lexicon Through the Use of Web-Based Material ..... 118

vii

5.1 Abstract ............................................................................................................................................... 118

5.2 Introduction ........................................................................................................................................ 119

5.3 Materials and Methods ...................................................................................................................... 122

5.4 Results and Discussion ...................................................................................................................... 123

5.5 Tables and Figure ............................................................................................................................... 129

5.6 References ........................................................................................................................................... 134

Chapter 6: Characterization of Sensory Differences in Premium Rums Through the Use of

Descriptive Analysis Panel ............................................................................................................................ 137

6.1 Abstract ............................................................................................................................................... 137

6.2 Introduction ........................................................................................................................................ 137



6.5 Tables and Figures ............................................................................................................................. 153

6.6 References ........................................................................................................................................... 165

Chapter 7: Effect of Ethanol Concentration on the Flavor Perception of Rum ................................. 167

7.1 Abstract ............................................................................................................................................... 167

7.2 Introduction ........................................................................................................................................ 167



7.5 Tables and Figures ............................................................................................................................. 179

7.6 References ........................................................................................................................................... 193

Chapter 8: Summary, Conclusions, and Future Research ........................................................................ 198

Appendix A: Standard curves ..................................................................................................................... 202

Appendix B: IRB approval letter for threshold determination in alcoholic matrices .......................... 236

Appendix C: Determination of odor thresholds in a 40% ABV ethanolic matrix ............................... 237

Appendix D: IRB approval letter for descriptive analysis panels ........................................................... 240

Appendix E: Rum DA panel recruitment questionnaire ......................................................................... 241

Appendix F: Rum DA panel prescreening ballot ...................................................................................... 244

Appendix G: Rum DA panel informed consent form ............................................................................. 245

Appendix H: Sample term generation sheet for descriptive analysis panels. ........................................ 247

Appendix I: Product information for attributes for rum descriptive analysis panel ............................ 248

Appendix J: Eigenvalues for factor loading using covariance matrix .................................................... 250

viii

Appendix K: Principal component analysis factor correlations of significant attributes on principal

component 1 (PC1) and 2 (PC2) for all nine rum samples in covariance matrix ................................. 251

Appendix L: Product information for attributes for ethanol dilution descriptive analysis panel ...... 252

Appendix M: Principal component analysis factor correlations of significant attributes on principal

component 1 (PC1) and 2 (PC2) for DR12 dilution samples in covariance matrix ............................ 254

Appendix N: Principal component analysis factor correlations of significant attributes on principal

component 1 (PC1) and 2 (PC2) for RA7 dilution samples in covariance matrix ............................... 255

Appendix O: Detailed day-by-day procedural description of rum descriptive analysis panel ............ 256

Appendix P: Detailed day-by-day procedural explanation of ethanol dilution descriptive analysis

panel ................................................................................................................................................................. 265

Appendix Q: Permission statements ........................................................................................................... 270

ix

List of Tables

Table 3.1 Product and manufacturing information for rums analyzed .................................................... 57

Table 3.2 Odor-active compounds identified by AEDA of nine rums ................................................... 58

Table 4.1 Compounds, isotopes, selected ions, and response factors used for stable isotope dilution

analysis ............................................................................................................................................................... 89

Table 4.2 Concentrations of the most important aroma compounds in nine rums .............................. 93

Table 4.3 ANOVA F-ratios for quantitation data ....................................................................................... 95

Table 4.4 Odor activity values and odor thresholds for most important compounds in rum ............. 97

Table 4.5 Pearson correlation coefficients for quantitation and sensory (aroma) data ....................... 104

Table 4.6 Pearson correlation coefficients for odor activity values and sensory (aroma) data .......... 106

Table 4.7 Pearson correlation coefficients for FD factors ≥ 8 and sensory (aroma) data .................. 108

Table 4.8 Pearson correlation coefficients for FD factors ≥ 16 and sensory (aroma) data ................ 110

Table 5.1 Web-based material used to create the rum flavor lexicon .................................................... 129

Table 5.2 Categorization of terms generated from web-material used to create the flavor wheel .... 131

Table 5.3 Top 15 descriptors for each category of rum ........................................................................... 132

Table 6.1 List of final attributes, definitions, references, reference scores, and reference preparations

determined for the descriptive analysis panel on rum .............................................................................. 153

Table 6.2 Analysis of variance (ANOVA) F-ratios for sensory attributes rated for nine rum samples ..

..................................................................................................................................................................... 156

Table 6.3 Mean intensity rating for significant aroma, mouthfeel, taste, aftertaste and aroma-by-

mouth attributes of the nine rum samples ................................................................................................. 157

Table 6.4 Pearson correlation coefficients for significant attributes for all nine rum samples .......... 163

Table 7.1 List of final attributes, definitions, references, reference scores, and reference preparations

determined for the descriptive analysis panel on ethanol’s effect on the flavor perception of

rums ................................................................................................................................................................ 179

Table 7.2 Analysis of variance (ANOVA) F-ratios for sensory attributes rated for dilutions of

Diplomatico Reserva 12-year rum ............................................................................................................... 181


mouth attributes of the Diplomatico Reserva 12 year rum dilution series ............................................ 182

Table 7.4 Pearson correlation coefficients for significant attributes for DR12 dilution series samples .

..................................................................................................................................................................... 185

x

Table 7.5 Analysis of variance (ANOVA) F-ratios for sensory attributes rated for dilutions of Ron

Abuelo 7 year rum ......................................................................................................................................... 187


mouth attributes of the Ron Abuelo 7 year rum dilution series ............................................................. 188

Table 7.7 Pearson correlation coefficients for significant attributes for RA7 dilution series

samples ............................................................................................................................................................ 191

xi

List of Figures

Figure 2.1 Traditional single pot distillation for rum production ............................................................. 32

Figure 2.2 Diagram of the solera aging system for rum ............................................................................. 33

Figure 3.1 EI-Mass spectrum of unknown 59 ............................................................................................. 61

Figure 4.1 Isotopically labeled standards used for quantitation ................................................................ 90

Figure 4.2 Cluster analysis of nine rums based on quantitation data ....................................................... 96

Figure 4.3 Cluster analysis of nine rums based on OAV data ................................................................... 99

Figure 4.4 Principal component analysis of quantitation and sensory data .......................................... 100

Figure 4.5 Principal component analysis of odor activity values and sensory data .............................. 101

Figure 4.6 Principal component analysis of flavor dilution factors ≥ 8 and sensory data .................. 102

Figure 4.7 Principal component analysis of flavor dilution factors ≥ 16 and sensory data ................ 103

Figure 5.1 Rum flavor wheel developed from web-based material describing white, gold, and aged

rums ................................................................................................................................................................. 133

Figure 6.1 Spider plots of mean significant attribute intensities for all nine rum samples ................. 158

Figure 6.2 Overlaid spider plot of mean significant attribute intensities for all nine rum samples ... 159

Figure 6.3 Spider plot of mean significant attribute intensities for seven rum samples, excluding

Diplomatica Reverva 12 year and Dictador XO Insolent ........................................................................ 160

Figure 6.4 Principal component analysis biplot of significant attributes present on principal

component 1 (PC1) and 2 (PC2) by the covariance matrix of mean significant attribute intensity

rating across all nine rum samples ............................................................................................................... 161

Figure 6.5 Cluster analysis constructed using mean significant attribute intensities for all nine rums ....

..................................................................................................................................................................... 162

Figure 7.1 Spider plot of mean significant attribute intensities the Diplomatico Reserva 12 year rum

dilution series .................................................................................................................................................. 183


component 1 (PC1) and 2 (PC2) by the correlation matrix of mean significant attribute intensity

rating across all three Diplomatico Reserva 12 year dilution samples ................................................... 184

Figure 7.3 Spider plot of mean significant attribute intensities the Ron Abuelo 7 year rum dilution

series ................................................................................................................................................................. 189

xii


component 1 (PC1) and 2 (PC2) by the correlation matrix of mean significant attribute intensity

rating across all three Ron Abuelo 7 year dilution samples ..................................................................... 190

1

Chapter 1: Introduction

For over 350 years, rum has been produced by the fermentation of sugar cane juice, syrup or

molasses, followed by distillation and then aging in oak barrels. Historically, rum production has

been associated with the Caribbean, which is still a major sugar cane and rum producing region.

Because it has a somewhat simple standard of identity, rum is a highly varied product, with the only

requirement being that it must be produced from sugar cane or its byproducts. The limited

definition allows for a breadth of product variety that is not typical of other spirit classes.

Classifications of rum include white, gold, aged, over-proof, light, heavy, industrial and Agricole.

Additionally, wide variation exists within these categories as well. The lack of a rigid standard of

identity and limited regulation allows for manufacturers to pick and choose from a variety of

manufacturing practices when they are creating their rum. Not only can rum be made from different

types of starting materials but variation also exists in types of yeast and bacteria used for

fermentation, length of fermentation, distillation apparatus and method used, barrel type and length

of maturation. As a result, currently over 1,500 individual rums are on the market, not including the

hundreds of flavored and spiced rums also being produced. In today’s drink and bar culture, rum is

experiencing a resurgence. High-quality rums, typically those aged at least five years and regarded as

best of their class, are being compared with fine spirits, such as Bourbon, Brandy, Cognac, and

Scotch. The average consumer is starting to regard rum as more than just a spirit to add to mixed

drinks such as daiquiris or mojitos.

Even with the recent cultural interest in rum, scientific research evaluating rum flavor as it pertains

to the category as a whole is lacking. Other distilled spirits, particularly Scotch and Irish whiskeys,

have research centers dedicated to the study of the complex chemistry of these beverages.

Numerous studies have been performed identifying the volatile compounds present in rum (Batiz &

Rosado, 1978; Bober & Haddaway, 1963; Leppänen, Denslow, & Ronkainen, 1979; Liebich, Koeing,

& Bayer, 1970; Maarse & tem Noever de Brauw, 1966; Ng, 1999; Pino et al., 2002; Pino, 2007; ter

Heide, Schaap, Wobben, de Valosis, & Timmer, 1981; Timmer, ter Heide, Wobben, & de Valois,

1971; Wobben, Timmer, ter Heide, & de Valois, 1971). These studies primarily focused on

identifying all of the instrumentally detectable volatile compounds and provided no indication of the

odor-activity or importance of the compounds identified to overall rum flavor. Only in the last

2

decade have studies begun to identify the odor-active compounds present in rum through the use of

gas chromatography-olfactometry (GC-O) (Burnside, 2012; de Souza, Vásquez, del Mastro, Acree, &

Lavin, 2006; Franitza, Granvogl, & Schieberle, 2016a, 2016b; Monsalve, Lopez, & Zapata, 2016;

Pino, Tolle, Gök, & Winterhalter, 2012). The limiting factor of these studies is that only one or two

rum samples were evaluated, with the exception of Monsalve’s evaluation of six Colombian rums,

and minimal if any product information was given for the rums, making it almost impossible to

repeat the studies. As a result, it is difficult to comment on the overall distinguishing attributes of

rum as a beverage class. This is especially important in the case of rum since production is highly

variable due to the simplicity of its standard of identity and diversity within the product class.

Therefore, flavor analysis of several different types of rums needs to be conducted.

Limited sensory analysis has been conducted on rums as well. Aroma profile analysis has been used

in a few studies to gain a basic understanding of the sensory attributes perceived in the samples and

was later used for comparison of the created models (Franitza et al., 2016a, 2016b). Sensory analysis

has also been used to compare rum and cachaça samples through descriptive analysis panels (de

Souza et al., 2006; Magnani, 2009). The most comprehensive study on rums was done by Gomez

(2002), who developed a preliminary lexicon for rum aroma followed by a descriptive analysis panel

evaluating nine rums. Rum is typically characterized as caramel, spicy (clove-like), fruity, and vanilla.

Sensory studies focusing on understanding rum as a class and the differences which exist between

and within the different categories still need to be performed. Additionally, being able to articulate

and describe the aroma perceptions experienced when drinking rum is important. This is essential

not only for the manufacturer to market their product but also for the consumer to be able to

verbalize what they are experiencing as they drink rum. Currently, no flavor wheel or lexicon has

been published for rum. Flavor wheels have been created for a variety of other distilled spirits

including brandy, cognac, and whiskey. While these wheels may contain many terms useful for

describing rums, they are missing certain attributes that are more nuanced and specific to rum.

Development of a flavor lexicon that encompasses many different types of rum would help to

articulate the flavor differences of rums within a single category as well as among different styles of

rum.

While rum flavor is largely driven by the odor-active constituents, the overall sensory perception can

also be affected by the alcohol concentration, as well as non-volatile compounds. Consumers drink

rum in a variety of ways including straight, “on the rocks” (with ice), or diluted with water. Similarly,

3

it is normal practice in the whiskey industry to dilute samples to 23% alcohol by volume (ABV)

before accessing the aroma for blending purposes. The given explanation for this practice is to

reduce the pungency of ethanol experienced at high alcohol concentrations. In addition to reducing

pungency, studies have also shown that changing the water/ethanol ratio also can significantly

impact the solubility and partition coefficients of the flavor compounds (Aznar, Tsachaki, Linforth,

Ferreira, & Taylor, 2004; Boothroyd, Linforth, & Cook, 2012; Taylor et al., 2010; Tsachaki et al.,

2008; Tsachaki, Aznar, Linforth, & Taylor, 2006; Tsachaki, Linforth, & Taylor, 2005). These changes

are dependent on if the beverage is evaluated in a static or dynamic system. This work has primarily

been done in model wine matrixes (12% ABV), and no studies have examined the effect of higher

ethanol concentrations on alcoholic systems under dynamic conditions. Nevertheless, none of this

work has been linked with sensory data to understand if consumers perceive changes in the overall

aroma as a result of changes in ethanol concentration. Establishing if there are perceivable sensory

changes caused by differences in ethanol concentration should be established before further

physiochemical interactions are investigated.

The central hypothesis of this study is that premium aged rums, while produced using a variety of

methods, still have a defining set of aroma characteristics caused by a unique combination of flavor

compounds that set these rums apart from those of lower quality, or so-called mixing rums, and that

ethanol concentration plays an important role in the perception of overall rum flavor. Therefore, the

goal of the study was to better understand the complex flavor chemistry of rum, including its aroma

composition and the effect of ethanol on flavor perception; with a main focus on premium aged

rums.

This objectives of this study were to 1) analyze a variety of premium aged rums, those consistently

rated by experts as being the best of the class, and several lower quality or mixing rums, to identify

the key aroma compounds, 2) quantitate the key aroma compounds in each of the nine rums and

compare the odorant compositions of the premium rums with those of lower quality or mixing

rums, 3) develop a rum flavor lexicon for sensory evaluation of premium aged rums, 4) evaluate the

sensory attributes of the nine rum samples to validate the rum lexicon as well as to correlate the

sensory attributes of the rums with the analytical findings from aims one and two, and 5) evaluate

the importance of ethanol concentration on the perception of flavor in premium aged rums through

sensory studies.

4

Rum is an important distilled spirit in the global alcoholic beverage market with an ever increasing

market share. The goal of this study was to better understand the flavor chemistry of rum products

as a whole, primarily premium rum, in terms or analytical and sensory evaluation. Increasing the

understanding of the flavor chemistry of rum products will help to better define rum as a category

and articulate the differences that exist between rum categories. Sensory evaluation of rum will begin

to establish how rums are perceived by consumers and a developed flavor lexicon will aid in

communication between all levels of rum production and consumers.

This research is novel in variety of ways as it is the first to 1) identify and quantify the odor-active

compounds for more than two rum samples in the same study, 2) focus on the difference between

mixing and premium rums, 3) create a flavor lexicon for rum, 4) develop a flavor lexicon using web-

based material rather than a traditional descriptive analysis panel, 5) evaluate the changes in sensory

perception as a results of changes in ethanol concentration, and 6) use chemometrics to correlate

sensory attributes with analytical analyses of rum.

5

1.1 References

Aznar, M., Tsachaki, M., Linforth, R. S. T., Ferreira, V., & Taylor, A. J. (2004). Headspace analysis of

volatile organic compounds from ethanolic systems by direct APCI-MS. International Journal of

Mass Spectrometry, 239, 17–25. http://doi.org/10.1016/j.ijms.2004.09.001

Batiz, H., & Rosado, E. (1978). A Single Column Gas Chromatographic Method for the Direct

Trace Analysis of High Boiling Components in Rums. The Journal of Agriculture of University of

Puerto Rico, 62(4), 330–343.

Bober, A., & Haddaway, L. W. (1963). Gas Chromatographic Identificaon of Alcoholic Beverages.

Journal of Gas Chromatography, 1(12), 8–13.

Boothroyd, E. L., Linforth, R. S. T., & Cook, D. J. (2012). Effects of Ethanol and Long-Chain Ethyl

Ester Concentrations on Volatile Partitioning in a Whisky Model System. Journal of Agricultural

and Food Chemistry, 60, 9959–9966. http://doi.org/10.1021/jf3022892

Burnside, E. E. (2012). Characterization of Volatiles in Commercial and Self-Prepared Rum Ethers and

Comparison with Key Aroma Compounds of Rum. University of Illinois at Urbana-Champaign.

Retrieved from https://www.ideals.illinois.edu/handle/2142/34368

de Souza, M. D. C. A., Vásquez, P., del Mastro, N. L., Acree, T. E., & Lavin, E. H. (2006).

Characterization of Cachaça and Rum Aroma. Journal of Agricultural and Food Chemistry, 54, 485–

488. http://doi.org/10.1021/jf0511190

Franitza, L., Granvogl, M., & Schieberle, P. (2016a). Characterization of the Key Aroma

Compounds in Two Commercial Rums by Means of the Sensomics Approach. Journal of

Agricultural and Food Chemistry, 64, 637–645. http://doi.org/10.1021/acs.jafc.5b05426

Franitza, L., Granvogl, M., & Schieberle, P. (2016b). Influence of the Production Process on the

Key Aroma Compounds of Rum: From Molasses to the Spirit. Journal of Agricultural and Food

Chemistry, 64, 9041–9053. http://doi.org/10.1021/acs.jafc.6b04046

Gómez, S. M. (2002). Rum Aroma Descriptive Analysis. Louisiana State University. Retrieved from

http://etd.lsu.edu/docs/available/etd-1114102-110301/

6

Leppänen, O., Denslow, J., & Ronkainen, P. (1979). A Gas Chromatographic Method for the

Accurate Determination of Low Concentration of Volatile Sulfur Compounds in Alcoholic

Beverages. Journal of the Institute of Brewing, 85, 350–353.

Liebich, H. M., Koeing, W. A., & Bayer, E. (1970). Analysis of the Flavor of Rum by Gas-Liquid

Chromatography and Mass Spectrometry. Journal of Chromatographic Science, 8, 527–533.

Retrieved from http://chromsci.oxfordjournals.org/content/8/9/527.short

Maarse, H., & tem Noever de Brauw, M. C. (1966). The Analysis of Volatile Components of Jamaica

Rum. Journal of Food Science, 31(6), 951–955. Retrieved from

http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2621.1966.tb03275.x/full

Magnani, B. D. (2009). Estudo Comparativo das Características Sensoriais do Rum e da Cachaça Estudo

Comparativo das Características Sensoriais do Rum e da Cachaça. Universidade Estadual Paulista.

Monsalve, J. O., Lopez, C., & Zapata, J. (2016). Characterization of aroma compounds in

Colombian rums by HS-SPME-GC-MS-O. Revista Colombiana de Química, 45(2), 48–54.

http://doi.org/http://dx.doi.org/10.15446/rev.colomb.quim.v45n2.60406.

Ng, L.-K. (1999). Analysis of Vodkas and White Rums by SPME-GC-MS. In J. Pawliszyn (Ed.),

Applications of Solid Phase Microextraction (pp. 393–406). Cambridge, UK: The Royal Society of

Chemistry. http://doi.org/10.1039/9781847550149-00393

Pino, J. A. (2007). Characterization of rum using solid-phase microextraction with gas

chromatography–mass spectrometry. Food Chemistry, 104, 421–428.

http://doi.org/10.1016/j.foodchem.2006.09.031

Pino, J. A., Tolle, S., Gök, R., & Winterhalter, P. (2012). Characterisation of odour-active

compounds in aged rum. Food Chemistry, 132, 1436–1441.


Pino, J., Martí, M. P., Mestres, M., Pérez, J., Busto, O., & Guasch, J. (2002). Headspace solid-phase

microextraction of higher fatty acid ethyl esters in white rum aroma. Journal of Chromatography A,

954, 51–57.

Taylor, A. J., Tsachaki, M., Lopez, R., Morris, C., Ferreira, V., & Wolf, B. (2010). Odorant Release

7

from Alcoholic Beverages. ACS Symposium Series, 1036, 161–175. http://doi.org/10.1021/bk-

2010-1036.ch012

ter Heide, R., Schaap, H., Wobben, H. J., de Valosis, P. J., & Timmer, R. (1981). Flavor Constituents

in Rum. In The Quality of Foods and Beverages (pp. 183–200).

Timmer, R., ter Heide, R., Wobben, H. J., & de Valois, P. J. (1971). Phenolic Compounds in Rum.

Journal of Food Science, 36, 462–463.

Tsachaki, M., Aznar, M., Linforth, R. S. T., & Taylor, A. J. (2006). Dynamics of flavour release from

ethanolic solutions. Developments in Food Science, 43, 441–444. http://doi.org/10.1016/S0167-

4501(06)80104-4

Tsachaki, M., Gady, A.-L., Kalopesas, M., Linforth, R. S. T., Athès, V., Marin, M., & Taylor, A. J.

(2008). Effect of Ethanol, Temperature, and Gas Flow Rate on Volatile Release from Aqueous

Solutions under Dynamic Headspace Dilution Conditions. Journal of Agricultural and Food

Chemistry, 56, 5308–5315. http://doi.org/10.1021/jf800225y

Tsachaki, M., Linforth, R. S. T., & Taylor, A. J. (2005). Dynamic Headspace Analysis of the Release

of Volatile Organic Compounds from Ethanolic Systems by Direct APCI-MS. Journal of

Agricultural and Food Chemistry, 53, 8328–8333. http://doi.org/10.1021/jf051202n

Wobben, H. J., Timmer, R., ter Heide, R., & de Valois, P. J. (1971). Nitrogen Compounds in Rum

and Whiskey. Journal of Food Science, 36, 464–465.

8

Chapter 2: Literature Review

2.1 Introduction

Rum and sugar cane have been major products and exports of the Caribbean since the region was

discovered by the Western world. Over the course of centuries, rum has evolved from the initial

harsh, unpleasant spirit to a sophisticated beverage that can be served neat similar to aged whiskeys

and Scotch. Even though rum has been studied since the early 20th century, literature about the

aroma of rum is limited.

2.2 History and Production of Rum

Rum is a distilled spirit made from sugar cane and commonly associated with the Caribbean. The

federal standard of identity for rum states:

“Rum” is an alcoholic distillate from the fermented juice of sugar cane, sugar cane syrup, sugar cane

molasses, or other sugar cane by-products, produced at less than 190° proof in such manner that the

distillate possesses the taste, aroma, and characteristics generally attributed to rum, and bottled at

not less than 80° proof; and also includes mixtures solely of such distillates. (Labeling and

advertising of distilled spirits, 27 C.F.R. § 5.22, 1969).

As can be seen from the definition above, there are no manufacturing regulations imposed on rum

other than it must come from sugar cane. This is different from other distilled spirits, which are

highly regulated. Bourbon for instance must be at least 51% corn, distilled to not more than 160°

proof, placed into the barrel for aging at no more than 125° proof and then aged for at least 2 years

in charred new oak barrels (Labeling and advertising of distilled spirits, 27 C.F.R. § 5.22, 1969). Rum

has none of these limitations. The spirit can be distilled by a variety of methods, aged in any type of

barrel, and the age label is not standardized but rather set by each country of production. As a result,

rum is a highly variable spirit.

9

Rums can be sorted into a number of categories including white, gold, aged, black, over-proof,

flavored, and spiced (Ayala, 2001; Fahrasmane, 2014). White rums are usually unaged or aged for

only a short amount of time in oak barrels or stainless steel vats and then filtered through charcoal

to remove any trace of color that may have developed during aging. Gold rums are typically aged for

1 to 3 years and have a light amber color. Aged rums are matured in wood barrels anywhere from 3

to 20 plus years and acquire a darker brown color due to the extended time in the barrel. Rums

within this category are those typically regarded as higher quality or premium rums. Black rums are

those that have excess caramel added to give them a darker color and therefore color is an indicator

that the rum has been aged for an extended period of time. Overproof rums are bottled at higher

alcohol contents, typically anywhere from 125° to 160° proof (62 - 80%ABV). Finally, flavored and

spiced rums have flavored extracts or fruits added to them after distillation and before bottling, or

have spices added prior to aging, respectively. Since fruits, flavors, and spices are added to change

the overall flavor of these rums, these products should not be compared with rums aged in the

traditional manner.

Rum can additionally be categorized based on the sugar cane product used for fermentation. When

sugar cane juice is the starting material, the final rum is labeled as rhum agricole (Ayala, 2001;

Fahrasmane, 2014). This is a production style that is usually found in the French Caribbean,

specifically the island of Martinique. If the rum is produced instead from molasses, it is categorized

as rhum industriel. This French term is typically not used for marketing purposes outside of the

French Isles (Fahrasmane, 2014).

Rums can also be categorized as light, heavy and traditional rums. Light rums are both light in color

and light in flavor and are produced using continuous distillation (Fahrasmane, 2014). Light rums

are traditionally produced on the islands of Cuba, the Dominican Republic, Haiti, Puerto Rico, and

the Virgin Islands (Jeffers, 1997). Heavy rums are those distilled using pot stills and fermented for a

longer period of time with both yeast and bacteria to produce a more complex and aromatic rum.

Jamaican rums are typically produced in this style (Fahrasmane, 2014). Traditional rums are those in

between, which have a medium flavor intensity (Fahrasmane, 2014).

Rum is an interesting and complex spirit, encompassing a variety of different flavor profiles and

production styles. Likewise, the history of the spirit may be as complex and vague as the spirit itself.

10

History of Sugar

It word be inadequate to discuss the history of rum without first making a brief mention of the

history of sugarcane, the main ingredient and standard of identity for rum. The origin of sugar cane

(Saccharum officinarum) goes back at least 10,000 years where it is thought to be first cultivated in New

Guinea (Foss, 2012; Macinnis, 2002). From there cultivation expanded to the other islands of East

Asia where it is still produced today. Sugar cane then traveled to India and is known to be cultivated

there by 500 B.C. as it is referenced in several texts (Macinnis, 2002). Sugar cane eventually made its

way to the Middle East although how and when this occurred is not documented. As sugar cane

became established in the Middle East, cultivation and production began to prosper (Williams,

2005). Sugar cane requires a lot of water, a warm and sunny environment to grow, and a large labor

force to process the cane into sugar. The Arabs were the first to introduce irrigation to sugar cane

production, which allowed for more continuous and even water to be delivered to the crop. Arabs

were also the first people to use slaves as the labor force for sugar cane production, and slave labor

would continue to be tied to sugar cane production until the late 1800’s.

During this time, the Muslims in the Middle East began trading sugar with Europeans. The first

written record of sugar trade with the West dates back to 95 AD (Foss, 2012). This was the begging

of a long history of Europeans trading with the Middle and Far East for sugar. It would not be until

the 1400’s that Europeans would decide to cultivate sugarcane on their own, constructing sugar

plantations in the colonies they were acquiring, instead of paying the high prices for purchasing

sugar from the East.

The first to do this were the Spanish, who first planted sugar cane on the Azore islands off the

western coast of Africa. This was a good place for sugar cane cultivation although the winds blowing

out to sea from the Sahara made it difficult to travel down the coast to transport the sugar cane.

From there as the European mother countries began to settle the islands as colonies in the

Caribbean sugar cane plantations began to spring up on those. Sugar cane was brought to the

Caribbean by Christopher Columbus on his second voyage to the West Indies in 1493 (Blue, 2004;

Macinnis, 2002). Sugar cane plantations began soon after, with records of plantations in the

Caribbean dating back to the early 1500’s (Foss, 2012). By the middle of the 1600’s sugarcane

plantations were prospering in the West Indies.

11

History of Rum

Although sugar cane was first cultivated in the Pacific Islands, it was not transformed into the

distilled beverage that we know today until some several thousand years later and half way around

the globe. One of the earliest mentions of a fermented beverage being produced from sugar cane is

recorded in the Indian Vedic texts around 2000 B.C. (Nicol, 2003). The two main alcoholic

beverages referred to were called ‘sidhu’ and ‘gaudi’, made from cane juice and molasses respectively.

While these are the first recorded beverages produced from sugar cane, they were much different

than the refined distilled spirit that we know today as rum. Distillation would not be discovered until

sugar reached the Middle East and the beverage would not make its final transformation into rum

until sugar began being cultivated in the Caribbean.

Early sugar cane plantations focused solely on sugar production and refinement. The major by-

product of the process, molasses, as essentially a waste product (Blue, 2004). It had no inherent

value to the plantation owners and was typically fed to animals in their feed or used as fertilizer. It

wasn’t until the discovery that molasses could be fermented and distilled into a potable beverage that

molasses was seen as a valuable material.

No one is quite sure when or where molasses was first converted into rum. The most likely story is

that a slave dipped a spoon or cup into a bucket of molasses that had been sitting out for several

weeks. While siting the molasses had fermented and mixed with water, becoming the first crude

form of rum (Blue, 2004). Rum was most likely a drink for the slaves when it was first produced,

and it is unlikely that the Plantation owners would have made mention of the crude spirit their slaves

were drinking before it has begun to be refined. The birthplace of rum is typically associated with

Barbados, even though this was probably not the first or only island in the Caribbean producing rum

at this time (Foss, 2012). This most likely because the earliest written mention of rum is in a 1651

letter written by a visitor to the island of Barbados. In his letter he mentions, “The chief fuddling

they make in the island is Rumbullion, alias Kill-Divil, and this is made of sugar canes distilled, a hot,

hellish, and terrible liquor” (Blue, 2004). This ‘terrible’ liquor would over time be refined into a spirit

desired by both the colonies and Motherlands. Early references to rum also include eau de vie de cannes

(Smith, 2005), guildive and taffia (Fahrasmane, 2014).

12

There is quite a bit of lore and uncertainty that also surrounds how rum got its name. Some claim

rum is derived from ‘brum’, a cane based drink that has been made by the Malays for thousands of

years (Blue, 2004). Some think that rum comes from the Dutch whose seaman had drinking glasses

called ‘rummers’. Another common thought is that rum is an abbreviation of the Latin word for

sugar, Saccharum (Blue, 2004; Mariani, 1983; Nicol, 2003). There is also some thought that the

work rum comes from the Spanish word for the spirit ‘ron’, as the Spanish were distilling a sugar

cane beverage in the Caribbean before the British set foot there (Blue, 2004; Nicol, 2003). However,

it is more likely that the Spanish and French adopted the English term rum and subsequently

translated it to ‘ron’ and ‘rhum’ (initially rome), respectively (Barty-King & Massel, 1983; Smith,

2005). The most likely theory is that rum is a shortening of the Devon word ‘Rumbullion’, which

means a great tumult (Barty-King & Massel, 1983; Blue, 2004; Mariani, 1983; F. H. Smith, 2005).

This has more merit as the first mention of rum, quoted above, used this term to describe the

beverage. By the 18th century the spirits produced from sugar cane and molasses were commonly

called rum (Nicol, 2003).

As rum was being produced and refined in the Caribbean, the main ingredient, molasses, was also

being exported to New England where rum production quickly became established (Foss, 2012).

Rum production in New England began in the mid 1600’s and there are records of several distilleries

having been built in Long Island and Boston by the 1660s. Because molasses was so cheap to

import, a gallon of molasses could be turned into a gallon of rum and sold for over five times the

initial price. As a result, rum became a staple beverage for the New England colonies. Comparing

New England rum to those produced in the West Indies, New England rum was by far the inferior

product. Due to the colder climate of New England the aging process took longer and so the rum

that was imported from the Caribbean was always of higher quality and higher value. Rum became

such a mainstay of the economy in the American colonies that it was even used in place of currency.

There are many records of carpenters and workers being paid rum as part of their salary as well as

records of rum being treated and bartered between colonists.

As a means to try and control the where colonies were buying their goods from, Great Britain

implemented several taxes under the Acts of Trade and Navigation. These Acts included the

Molasses Act of 1733, the Gin Act of 1736, and the Sugar Act of 1764 which placed taxes on

products such as sugar, molasses and rum production. This was done to ensure that the colonists

bought their supplies from British merchants or stopped producing rum, which lead to smuggling of

13

molasses from Spanish and French merchants (Nicol, 2003). It is thought that the taxation of

molasses and sugar for rum production is one of the indirect causes of the American Revolution.

The New England colonists were also instrumental in the formation of the Triangle Trade. Sugar

production is a labor intensive process and requires a large work force to plant, maintain and harvest

the sugar cane. In the early days, English slaves were the predominant workforce on the plantations

but by the late 1600’s African slaves were the majority of plantation workers (Gately, 2008). These

slaves were brought to the Caribbean through an economic trade system known as the Triangle

Trade. Molasses and sugar produced in the Caribbean islands was traded to New England and in

New England the molasses was made into rum. The rum was then shipped to West Africa to trade

for African slaves who were then shipped to the Caribbean to work on the sugar cane plantations

(Foss, 2012). Britain was also a key player in the Triangle Trade, with goods from the Caribbean

being shipped to Liverpool and Bristol rather than New England. The triangle trade continued in

the Caribbean until the 1800’s when slavery started to be abolished.

Moving into the 19th century, rum production and popularity declined due to many factors.

Contributing to this was the Civil War, the creation of bourbon, the abolishment of rum rations in

the Navies (U.S. and Britain) and prohibition. Rum, however, has been recently experiencing a

renaissance. The introduction of Bacardi and Captain Morgan are responsible for helping to increase

the popularity of rum (Delavante, 2004). Today some premium rums are seen as being on par with

traditional top-shelf spirits such as Scotch and whiskey (Miles, 2015; Padgett, 2017). The recent

interest in rum is evident by the publications of several new books on the topic including Rum –

The Manual (Broom, 2016), Rum Curious: The Indispensable Tasting Guide to the World's Spirit

(Minnick, 2017), and The Curious Bartender's Rum Revolution (Stephenson, 2017).

Current Rum Market

A rum renaissance has been claimed for the past several years (Eberle, 2014; Freedman, 2016; Irani,

2013; Padgett, 2014, 2017; Ward, 2015). Premium rums are commonly compared with other high-

end distilled beverages such as Scotch and cognac (Fahrasmane, 2014; Freedman, 2016; Miles, 2015;

Padgett, 2017). While growth has occurred in the rum market, it has not seen the explosive growth

that was expected (Ward, 2015). However, experts in the field still predict growth in the rum market

in the coming years (Shoup, 2017).

14

From 2000 to 2010 the rum category increased 40% (by volume sales) (Hopkins, 2015). However

since 2012 rum has seen a slight decline (Shoup, 2017). From 2014 to 2015 rum sales declined 0.8%.

Additionally, only 39% of rum sales were premium products, meaning most consumers are

purchasing lower quality mixing rums (“Global Rum Insights - Market Forecasts, Product

Innovation and Consumer Trends,” 2016). In the United States total rum consumption declined

1.5% (by sales volume) during the same time period, while premium rums rose 2.8% (in volume)

(Swartz, 2016).

As of 2015, rum comprised 5.6% of the global spirit’s market (by value) (Global Spirits, 2016). In the

United States, rum makes up 13.7% of the market value, trailing vodka (30.5%) and whiskey (23.6%)

(Spirits in the United States, 2016).

Rum is still predicted to grow over the next several years particularly in the premium rum category

(Business Monitor International Research, 2016). One of the factors expected to drive this trend is

that global rum manufacturers are expected to increase their portfolio to include premium rums.

Nevertheless, other analysists predict that the global rum market will decline in the next five years.

The Rum Global Insights group predicts that rum consumption will drop 0.5% per year (“Global

Rum Market Set to Decline Over the Next Five Years,” 2017). However declines are mostly

expected for mixing rums, and premium rum sales are still expected to increase (“Global Rum

Market Set to Decline Over the Next Five Years,” 2017).

Currently, the top 5 rum brands worldwide are Mc.Dowell’s No.1 Celebration, Bacardi, Tanduay,

Captain Morgan and Havana Club (“Global Rum Market: Leading Brands Based on Sales Volume

2015,” 2017). Additionally, the top rum market in the world is India (404.2 million liters), followed

by the United States (241.9 million liters), the Philippines (169.7 million liters), the Dominican

Republic (54.7 million liters) and France (36 million liters) according to 2012 data (Hopkins, 2014).

Overall, it is difficult to predict what consumers will like and when their preference in their distilled

spirit of choice will change. Rum is noted as an underexploited category (Hopkins, 2015; Ward,

2015), while rum’s versatility is commonly referred to a strength of the category, suggesting lots of

room for growth (“Global Rum Insights - Market Forecasts, Product Innovation and Consumer

Trends,” 2016; Ward, 2015). The increase interest and trend for brown spirits, particularly whiskey

also bodes well for rum, exposing people to the realm craft spirits that are available (Ward, 2015).

15

Additionally, as the price of premium whisk(e)y continues to rise, consumers may turn to rum as a

cheaper alternative of similar quality (Eberle, 2014; Freedman, 2016; Shoup, 2017).

2.3 Rum Production

There are many factors which affect the final flavor of the rum. These include things such as sugar

cane variety, maturity of the sugar cane when harvested, the proof of the molasses used in the

fermentation as well as the type of yeast, the type of distillation performed, the type of barrel that

the final spirit is aged in and how the final product is blended.

The first step in the production of rum is the collection of the raw materials. Rum is typically

produced from molasses, a by-product of sugar production. Other forms of sugar cane can be used

to produce rum such as cane syrup or sugar cane juice.

Sugar Production

When the sugar cane is ready to be harvested, the cane fields are set on fire to sterilize the soil and

help limit moisture loss once the grass has been cut. The sugar cane can be harvested mechanically,

but it is usually done by hand, an extremely labor intensive process, as it is thought to produce a

superior final product. The sugar cane is then transported to the mill by truck or canal. Most sugar

cane is processed within 24 hours after harvesting to minimize moisture loss and conversion of the

sucrose to glucose and fructose by invertase or into a polysaccharide, dextran, by the soil

contaminant Leuconostoc mesenteroides (Nicol, 2003). When the cane reaches the factory, the leafy tops

are removed and the stalks are then cut and milled to release the can juice. Every hundred tons of

sugar cane harvested will yield ten tons of cane juice with a 10% sucrose w/w concentration. This

cane juice is the first by-product of sugar cane that can be used for rum production.

The cane juice will then move through a series of evaporators producing cane syrup which can also

be used in rum production (Nicol, 2003). The syrup is then left to crystallize and is finally

centrifuged to separate the sugar crystals from the remaining syrup. The crystals will be further

refined and sold as sugar, and the syrup left over once all the sucrose has crystallized out is known as

blackstrap molasses. This is the starting material for most rums. Many factors contribute to molasses

16

quality, including unfermentables, gums, nitrogen, and sulfur, with sugar content being the most

important.

Fermentation

Traditional rum production relied on wild fermentation, using yeasts and bacteria present in the air

and naturally occurring in the molasses. Today most fermentations are done using pure cultures of

yeast. The most common yeast used in rum fermentation is Saccharomyces cerevisiae, although

Saccharomyces bayanus and Schizosacharomyces pombe are also used (Nicol, 2003). Dunder can also be

employed to increase the production of volatiles during fermentation, as it consists of wild yeasts

and anaerobic bacteria. The dunder is the residue from wash distillations which are left to ferment in

a dunder pit before being added to the fermentation vat or the wash before distillation (Rogers,

2014). The bacteria that is present in the dunder produce volatile compounds that would not be

formed during a yeast only fermentation. The use of dunder in combination with Schizosacharomyces

yeast is typically used for the production of heavy rums (Fahrasmane, 2014).

The molasses is added to a fermentation tank and diluted to 16 - 20 degrees Brix for fermentation

(Nicol, 2003). The molasses needs to be diluted to reach the optimal sugar concentration for yeast

to ferment effectively without being stressed or having too much sugar available. During the filling

of the tank, the yeast inoculum is also added. The temperature of the vat is closely monitored and

maintained at 30 – 33°C. Fermentations left on their own would approach temperatures over 40 °C,

which would kill the yeast and hinder the alcohol production. The fermentations tanks are typically

cooled by circulating water through heat exchangers (Nicol, 2003). During the fermentation process,

the yeast convert glucose into ethanol and carbon dioxide.

During the initial fermentation (first 24 hours) a wash of 5-7% alcohol is produced. While

fermentation can be completed in as few as 24 hours, darker rums can be fermented for up to three

weeks (Blue, 2004). The shorter fermentation results in a lighter rum and longer fermentations

produce heavier rums that have higher concentrations of esters and other congeners.

17

Distillation

The wash is then distilled, increasing the alcohol concentration from 7% to 80 -94% ABV. There are

two main types of distillation apparatus used in rum production. These include pot and continuous

distillation.

The using of pot stills is the traditional method of rum distillation. This can be a single pot system

(Figure 2.1) or a double pot distillation system where two pot stills are connected in tandem. In this

technique the rum is added to a large wooden vat or copper pot, above which sits a copper

shoulders and swan neck which empties into a retort (Nicol, 2003). One of the reasons copper has

been used for stills is that it helps to remove off flavors by binding sulfur compounds (Fahrasmane,

2014). On top of the retort is a rectifier, which is connected through another tube to the condenser

(Nicol, 2003). Some simpler designs do not include the retort and rectifier portion.

At the beginning of the distillation, the pot is filled with ~6,000 L of wash and the retort is filled

with low wines remaining from the previous distillation at ~51-52% ABV (Nicol, 2003). The still is

heated by either steam or bagasse, the unusable portion of the sugarcane stalk. As the ethanol and

volatiles evaporate from the pot, they then travel through the swan neck into the retort where the

vapors collect and are known as low wines. The ethanol will then evaporate from the retort and pass

through the rectifier, essentially a condenser consisting of a container of water held at 45-50°C

which copper tubes pass through carrying the rum vapors. When the ethanol and volatiles leave the

rectifier they enter the condenser, from which the rum is collected at ~85% ABV. Rum is collected

from the condenser until the distillate reaches 43%ABV, at which point it is directed back to the low

wines. The low wines collected from the distillation are then used during the next distillation. Pot

distillation is slow process and the extended time the rum spends being heated for distillation also

allows for the reactions and formation of aroma compounds (Fahrasmane, 2014).

The other and more common technique is continuous distillation. These can be both single or

multiple column systems, with single column distillations used for the production of traditional rums

and multiple column distillations used for lighter body rums (Fahrasmane, 2014). Continuous

distillation is capable of producing ten times more rum than a traditional pot still and is therefore

favored by larger companies that receive molasses from several sugar manufacturers (Nicol, 2003).

18

Depending on the type of still used for distillation, different final products can be achieved. The pot

still is known for producing heavier rum with a higher amount of congeners. Column stills, on the

other hand, have the ability to alter which fractions are collected for the final product and can

therefore alter the level of various volatiles present in the final beverage.

Maturation

The aging of rums is one of the most important aspects of production. The length of time spent in

the cask dictates the type of rum that is being produce, with white rums only being aged for a couple

of years or less, gold rum for 2-5 years and aged rums for 5–20+ years. The type of barrel used will

greatly influence the final flavor profile. Rum, unlike other aged spirits such as Bourbon which has

strict laws regulating the types of barrels that can be used, can be aged in any type of barrel that the

manufacture chooses (Foss, 2012). This can include new oak barrels, used barrels that have been

previously used for beverages such as sherry, wine, whiskey, bourbon, cognac, etc. Additionally,

barrels that have been used several times may be better suited for light rums while newer casks

should be used for aged rums (Nicol, 2003).

The barrels are typically filled with rum at 83-85%ABV. Once filled, the barrels will be stored in a

warehouse until the desired organoleptic qualities have been achieved. Some rums will be transferred

to several casks during maturation, each imparting a different flavor profile to the final beverage.

Other rums may be blended after several months, and then the blend will continue aging until ready

to be bottled.

The solera aging system is another way to blend and age the rum (Figure 2.2). This style of

maturation is typically associated with the aging of sherry and thought to produce consistent quality

and character (Gonzáles Gordon, 1990; Reader & Dominguez, 2003). Casks on the bottom or solera

level, are those that hold the oldest rums. These casks are partially emptied, and the rum removed is

sent to bottling. Rum from the previous level or criaderas is removed to fill the solera level. The rum

removed from the barrels is first transferred to a tank where the rum from all the casks is blended

before being added to the next level. This continues for each criaderas, and then the youngest

criaderas is then filled with new rum. The amount of rum removed from each level, the frequency of

transfer between levels, and the number of criaderas is dependent on the producer. The levels do

not need to be stacked on top of each other and may even be stored in different warehouses.

19

In terms of maturation, it is also important to consider the climate that rums are aged in. Unlike

many of the distilled spirits made and aged in America or Europe which is a considerably cooler

climate, rum is aged in the hot tropical environment of the Caribbean. The hot and humid climate

(27-32°C and 75-90% humidity consistently) increases the speed of maturation (Foss, 2012; Nicol,

2003). Therefore if the same spirit was aged it in Scotland and the Caribbean, the aging process

would progress much quicker in the Caribbean making it easier to achieve a well-aged rum in a

shorter amount of time. The downside of this is that because of the hotter climate and increased

humidity there is a larger loss of rum to evaporation over time.

During the maturation process, the rum interacts with both itself and the barrel it is being aged in

(Fahrasmane, 2014). First, wood constituents are extracted from the barrel into the rum.

Additionally, the lignin extracted from the barrel as well as alcohols present in the rum can undergo

oxidation. And finally, new volatile compounds can form from the interactions of various

compounds in rum.

Blending and Bottling

When maturation is completed, the rum will be dumped from the cask and transferred to the

bottling facility. Along the way, the rum will be diluted to its final strength of 40-43% ABV. White

rum will be sent through charcoal filtration systems before bottling to remove any color that may

have developed during aging. This process can also remove some of the volatiles from the spirit as

well.

Almost all rums that are produced and sold are blends. It is very uncommon to see a rum which is

from a single cask. Most rums are a blended from numerous casks which consists of rums which

have aged for different amounts of time (Blue, 2004). The blending of rum allows for a consistent

product that is reproducible, something that is not achievable in a single barrel. Most distilleries have

a master blender who is in charge of creating and maintaining the specific blend for a brand or line

of rums.

20

2.4 Flavor Research on Rum

While rum has been produced since the 1600’s, it was not until the early twentieth century that rum

flavor started to be analyzed scientifically. Although rum has been studied academically for the past

70+ years, there is still a relatively small body of literature regarding the flavor characteristics of rum.

Rum is first mentioned in the 1939 paper entitled “The Aroma of Rum” (Arroyo, 1939), in which

the author discusses the general composition of rum. Even though rum had been produced for

hundreds of years by this time, the author makes it a point to note that they understand the difficulty

in producing rum and that creating a consistent product even within the same factory was very

difficult. However, they were beginning to gain a better understanding of the impact of processing

on the final aroma. Arroyo notes that the two most important factors contributing to the final rum

aroma are the compounds formed during fermentation, which remain after distillation, and those

formed during the maturation process. A limiting factor in their ability to understand the aroma of

rum was the primitive analytical techniques available at the time. At this point in time, only classes of

compounds could be distinguished, such as esters, alcohols, and phenolics and not individual

components.

As rum research moved into the 1960’s and 1970’s, gas-chromatography became more prevalent and

allowed researchers to identify for the first time the individual components of rum. There was

extensive research during this time into understanding the volatile composition of rum, but there are

many limitations on this research and how it can be applied today.

Initial research was performed by the U.S. Customs laboratory, trying to determine how to

distinguish authentic spirits from those that had been doctored or counterfeited (Bober &

Haddaway, 1963). Gas chromatography paired with a flame ionization detector (FID) was used to

compare chromatograms of different products. This technique also allowed them to quantify a small

number of compounds including n-propyl alcohol, isobutyl alcohol, d-amyl alcohol, isoamyl alcohol,

n-butyl alcohol, furfural, ethyl hexanoate, and ethyl octanoate. However, many of the peaks on the

chromatogram could not be identified at that time.

It was not until 1966 and introduction of gas chromatography paired with mass spectrometry that

many of the unknown compounds in rum were first identified (Maarse & tem Noever de Brauw,

1966). Maarse & tem Noever de Brauw examined the composition of Jamaica Rum and were able to

21

identify over 50 unique compounds. In order to identify the volatiles present in these samples, the

rum was extracted with pentane-ether, concentrated, and then fractionated by running the extract

through silica gel. Each fraction was then analyzed for the individual compounds using techniques

such as mass spectrometry, nuclear magnetic resonance, infrared spectroscopy, and comparison to

known standards (Liebich, Koeing, & Bayer, 1970; Maarse & tem Noever de Brauw, 1966).

During this time, a number of studies focused on identifying a key group of compounds present in

rum. These studies included identifications of ethyl esters of fatty acids (Stevens & Martin, 1965),

phenolic compounds (Timmer, ter Heide, Wobben, & de Valois, 1971), nitrogen compounds

(Wobben, Timmer, ter Heide, & de Valois, 1971), high boiling constituents (Batiz & Rosado, 1978),

and sulfur compounds (Leppänen, Denslow, & Ronkainen, 1979). Additionally, ter Heide and

colleagues attempted to identify all of the volatile compounds present in rum (ter Heide, Schaap,

Wobben, de Valosis, & Timmer, 1981). This lead to the identification of over 400 compounds, 214

of which were identified for the first time in rum. While this study shows the complexity of rum

flavor, the main problem with this research is that the volatile components were identified regardless

of the whether they were odor active or not. Additionally, compounds with low odor thresholds that

would have an impact on the aroma but would not be present in high enough concentration to be

detected were missed by this method of identification.

There are a number of other disadvantages to these studies that make the data collected difficult to

use in a meaningful way. First, technology has significantly advanced in the last 40 year, both in

terms of more sensitive and robust instrumentation, as well as overall knowledge of extraction

techniques and understanding of flavor in general. Additionally, many of the early methods created

artifacts, compounds that are not present in the actual samples but are formed during the extraction

or analysis. A huge disadvantage to studying the aroma chemistry of rum is that there was no way to

identify which of the volatile components where odor-active. None of the studies report odors for

individual compounds. A common practice of the time was to smell the fractions for aroma. This

only informed the scientist that a compound in that fraction was odor-active but they were not able

to determine which specific compounds were responsible for the odors. Furthermore, there is no

way to rank how individual compounds contribute to the overall aroma of rum.

As flavor research moved into the 1990’s, rum analysis saw the introduction of solid phase

microextraction (SPME) (Ng, 1999). Direct analysis of liquid samples as well as headspace sampling

22

can be performed with SPME. The use of SPME for sample analysis provided a number of

advantages. First, pre-concentration of the samples can be performed without the need of solvent.

This helped to reduce the number of artifacts formed, although new artifacts from the fiber were

present. Additionally, minimal sample preparation was required saving time previously spent for

sample extraction and concentration. However, SPME also has several limitations. Several different

fibers are available for SPME, and each one has a different affinity for different compounds. This

requires advanced knowledge of your target compounds to be able to choose the best fiber. Also,

SPME can over represent certain compounds, particularly smaller and more volatile ones, that have

a greater affinity for the fiber than heavier less volatile larger molecules. Furthermore, the extraction

time is very important to get an accurate representation of the volatiles present. Longer extraction

times allow more time for exchange or displacement among compounds absorbed in the fiber.

Additional research on rums using SPME was done by Jorge Pino. His first paper used SPME for

the extraction and quantification of ethyl esters in seven white rums (Pino et al., 2002). Extensive

method development was performed to accurately quantify the esters in alcoholic matrices,

determining that diluting the samples to 12% ABV along with the addition of salt to the sample

provided the best results. Ethyl hexanoate, ethyl octanoate, ethyl decanoate and ethyl dodecanoate

were quantified, and the results showed that ethyl decanoate was present in highest concentration in

every rum followed by ethyl octanoate, and ethyl dodecanoate. Ethyl hexanoate had the lowest

concentration of the esters analyzed in every rum sample. Recently, Pino has validated this method

for analysis of ethyl esters as well as the whiskey lactones in white rums (Pino & Roncal, 2016). Pino

then went on to use SPME to characterize the volatiles present in six samples of three and seven

year old rums (Pino, 2007). Pino was able to identify 184 different volatile compounds, but no

quantification was performed other than comparing peak areas.

It was not until the past decade that gas-chromatography olfactometry (GCO) was first utilized in

rum flavor research to identify the odor-active volatile compounds present in the beverage despite

the fact that the technique was introduced in the 1960’s (Acree & Barnard, 1994). The first GCO

study compared the aroma of cachaça, a Brazillian distilled spirit also produced from sugar cane, and

an unidentified Bacardi rum (de Souza, Vásquez, del Mastro, Acree, & Lavin, 2006). Researchers

used GCO on extracts of each spirit and compared the results to those obtained by descriptive

sensory analysis for the same samples. The top twenty odor-active compounds were quantified and

ranked according to their odor intensities. While this was the first study to identify the odor-active

23

compounds in rum, no quantitation of the compounds was performed. Rum was only compared to

cachaça by sensory evaluation and comparison of odor spectrum values (OSV), which is similar to

flavor dilution (FD) factors. β-Damascenone, diethyl acetal, ethyl 2-methylbutyrate, ethyl isobutyrate

and an unknown compound (RI 866 on DB5 column), were found to be the five most potent aroma

compounds in rum.

A second GCO study was conducted by Pino (Pino, Tolle, Gök, & Winterhalter, 2012). He

identified and quantified potent odorants that had a FD factor of 32 or higher. Compounds noted as

being important to the aroma of rum include diethyl acetal, ethyl 2-methylpropanoate, ethyl

butyrate, ethyl 2-methylbutanoate, 3-methylbutyl acetate, ethyl hexanoate, ethyl octanoate, ethyl

decanoate, 2-phenethyl acetate, β-damascenone, 2-phenylethanol, 2-methoxyphenol, 4-ethylguaiacol,

4-propyl-guaiacol, trans-whiskey lactone, cis-whiskey lactone, γ-nonalactone, eugenol and vanillin.

This was the first study to quantify the odor-active volatiles present in rum. Quantification was done

by internal standard methodology but using methyl octanoate. Additionally, this paper identified all

of the volatile compounds found in the rum sample. No unknowns were listed suggesting a lack of

verification with authentic standards.

Recent work done by the Cadwallader lab evaluated the similarity between rum and rum ethers

(Burnside, 2012). The study identified the most odor-active compounds present in Bacardi white

and gold rums, as well as in nine commercial and self-prepared rum ethers. Several compounds were

found to be essential to both rum and rum ether but both contained their own individually

important compounds. Forty-four compounds were identified in the two rum samples. AEDA was

used to determine the most potent odorants in the rums: ethyl propanoate, ethyl isobutyrate, isoamyl

alcohol, acetic acid, β-damascenone, phenethyl alcohol, cis-whiskey lactone, and vanillin. No

quantification was performed on the samples.

Another recent study evaluated nine Columbian rums by headspace SPME coupled with GC-MS-O

(Monsalve, López, & Zapata, 2016). Forty-six compounds were identified between the nine samples.

The results indicated that some compounds were able to be detected in all rums while others were

only present in some of the samples. No quantification of the identified compounds was conducted.

The two most thorough studies on rum were recently done by Franitza (Franitza, Granvogl, &

Schieberle, 2016a, 2016b). The first study evaluated two rum products, a solera aged rum produced

24

from molasses (rum A) and a lower quality rum readily purchased from a local store (Rum B)

(Franitza et al., 2016a). Samples were analyzed by GCO and AEDA leading to the identification of

40 odor-active compounds in rum A and 26 in rum B with FD factors between 8 and 2048. Overall,

a total of forty-five odorants were identified and quantified by stable isotope dilution analysis

(SIDA) in both rums. Odor activity values we also calculated for all compounds to gain a better

understanding of their overall impact on rum flavor. Ethyl 2-methylbutyrate, β-damascenone, 3-

methylbutanal, 2,3-butanedione, ethyl butyrate, 1,1-diethoxyethane, ethyl 3-methylbutanoate, ethyl

pentanoate, and ethyl hexanoate were all found to have OAV’s greater than 1 (indicating they are

present in concentrations above their odor threshold in the sample) in both rums. Several

compounds had OAV’s above 1 in just one of the samples including vanillin, cis-whiskey lactone,

eugenol, and guaiacol for rum A and 2-methylbutanal and 4-propylguaiacol for rum B.

Their second study evaluated how the flavor constituents of rum change throughout the production

process (Franitza et al., 2016b). The study specifically focused on the flavor of the molasses,

fermentation mash, distillation and final product. Forty-five compounds were identified in at least

one of the four production stages. The concentration of all volatile compounds increased during the

fermentation of the molasses except butanoic acid, phenylacetic acid, vanillin and (R)-2-methyl-1-

butanol which was only detected in the molasses [ (S)-2-methyl-1-butanol was found in all samples].

During the distillation step, the concentration of most volatiles again increased except acetic acid,

butanoic acid, phenylacetic acid, vanillin, 3-methylbutanoic acid, (R)- & (S)-2-methyl butanoic acid,

3-(methylthio)propanol, decanoic acid, 3-hydroxy-4,5-dimethylfuran-2(5H)-one, 2-phenylethanol, 4-

ethylphenol, and 4-methylphenol. During the aging of the rum, many of the volatile compounds

decreased or stayed at the same concentration. Compounds that increased during aging were acetic

acid, phenylacetic acid, vanillin, 3-(methylthio)propanal, decanoic acid, 3-hydroxy-4,5-dimethylfuran-

2-(5H)-one, 2-methoxyphenol, 4-ethylphenol, 4-methylphenol, 4-ethylguaiacol, 4-propylguaiacol,

and ethyl 3-methtylbutanoate, suggesting these compounds were either extracted from the wood

barrels or formed from reactions that occurred between compounds in the rum over time. Cis- and

trans-whiskey lactones were the only two compounds detected in the final aged product only,

verifying that they are extracted from the wooden barrels during aging.

Even though rum has been studied for almost 100 years, there is still much to learn in terms of the

flavor chemistry of rum. Studies that have identified the aroma-active compounds only look at one

or two rum samples. Since rum is such as highly variable spirit due to the different and numerous

25

manufacturing practices employed, the current body of literature is not able to completely describe

what defines rum as a class. Studies that survey a variety of rum types and manufacturing practices

are needed for a more thorough understanding of the aroma of rum.

2.5 Sensory Studies on Rum

A minimal amount of work has been performed to define the sensory characteristics of rum. A

master’s thesis by Gómez (2002), was the first sensory study to analyze the aroma profile of rums.

Gomez used a descriptive analysis panel to develop a flavor lexicon consisting of 33 aroma

attributes, developed from the evaluation of 15 rums. The second part or the research used a

descriptive analysis panel to describe the aroma differences for nine rums that varied in production

methods. Even though a lexicon had been constructed in the initial work, it was not adequate to

describe all of the differences perceived by the panel in the second study. Of the 22 terms evaluated,

seven were not found in the initial lexicon. This suggests that the rums chosen for the creation of

the lexicon did not adequately cover all product variations. The rum samples were able to be

differentiated with the main aroma differences in woody, buttery, caramel, honey, vanilla, cinnamon,

artificial fruit, cardboard, and ocean-like attributes.

A descriptive analysis panels have also been conducted to identify the sensory differences between

rum and cachaça (de Souza et al., 2006; Magnani, 2009). The first study by de Souza (2006)

compared a Joao Mendes cachaça to a Bacardi rum (only brands were specified, not specific

products). Ten aroma attributes were evaluated, and rum was found to be higher in caramel and

vanilla notes, while caramel was higher in grassy, sulfury, vinegar, spicy, citrus, alcohol, and melon

notes. Another descriptive analysis panel was performed by Magnani (2009) who evaluated two

rums and two cachaça samples for aroma and in-mouth sensory perceptions for seventeen

attributes. The results showed the cachaça and rum differed in terms of golden color, body

appearance, turbidity, wood aroma, wood flavor, sweet taste, and viscosity.

Several studies have characterized the aroma of rum through the aroma profile method (Franitza et

al., 2016a, 2016b). Both studies characterized rum as ethanolic, malty, butter-like, fruity, clove-like

and vanilla-like. The second study added the terms baked apple-like and caramel like (Franitza et al.,

2016b).

26

Overall, rum has been consistently described as caramel, spicy (clove-like), fruity and vanilla. While

several studies have evaluated the sensory profiles of rum, the significant variation in rum products

necessitates additional sensory studies on rum flavor.

2.6 Flavor Wheels

Lexicons and flavor wheels are created and used to provide a standardized vocabulary for enhanced

communication and discussion between sensory scientists, product developers, business clients and

consumers. Lexicons have been developed for a variety of different food products including

whiskey, wine, beer, brandy, cognac spices, cheese, bread, olive oil, almonds, tea and orange juice

(Capone, Tufariello, & Siciliano, 2013; Civille, Lapsley, Huang, Yada, & Seltsam, 2010; Drake,

McInvale, Gerard, Cadwallader, & Civille, 2001; Jolly & Hattingh, 2001; Kleinert, Bongartz, Raemy,

& Wadenswil, 2009; Koch, Muller, Joubert, van der Rijst, & Næs, 2012; Lawless, Hottenstein, &

Ellingsworth, 2012; Lee, Paterson, Piggott, & Richardson, 2001; Lurton, Ferrari, & Snakkers, 2012;

Mojet & de Jong, 1994; Pérez-Cacho, Galán-Soldevilla, Mahattanatawee, Elston, & Rouseff, 2008;

Schmelzle, 2009). The wine flavor wheel is the most well-known, recognized by coinsurers and

consumers alike. Having a set of standard terms, definitions, and references for the product

descriptors allows users to ensure that they are describing the same aspect of the product.

Lexicons are typically developed by a team of trained sensory panelists. Panelists usually have

hundreds of hours of training evaluating different food products performing descriptive analysis. To

develop a complete lexicon, a variety of different products that encompasses all of the different

aspects of a product category need to be evaluated. Taking the cheddar cheese lexicon as an

example, samples were chosen to comprise cheese from different geographical regions, producers,

and a range cheese maturity to develop the most comprehensive lexicon as possible (Drake et al.,

2001). Once the products to be evaluated have been selected panelists will generate terms to

describe their perception, determine appropriate chemical or food stuff references, developing

specific definitions for each attribute. After the initial term generation, any terms that are redundant

are removed from the list. Panelists will then practice scaling the attributes according to the selected

reference. Each sample will then be evaluated by the panelists with each individual attribute being

27

rated in comparison to the reference. The compiled data allows researchers to identify how the

products within the category differ from each other.

Once a lexicon has been developed, it can be converted into a flavor wheel, which provides a visual

representation of the generated terms. Terms are typically grouped by category, with the grouping

identified in the inner circle and the specific term listed along the outside. The completed wheel can

then be used to train new panelists as an aid to help with term selection and be able to communicate

with other scientists or consumers.

Rum is an extremely complex product due to its limited standard of identity. Since rum can come

from a variety of sugar cane by-products, be distilled in multiple ways and aged in any type of barrel

the manufacturer desires, there is a significant amount of variation between the products classified as

rum. There is currently no rum flavor wheel or lexicon available. Developing a lexicon would aid

both manufacturers and consumers in being able to better understand and describe the products.

2.7 Effects of Ethanol on Flavor Perception of Alcoholic Beverages

Ethanol is the major component of alcoholic beverages, other than water, but relatively little is

known about its effects on flavor perception. The alcoholic beverages industry is a $189 billion

industry, comprising 56% of the total beverage industry as of 2011 (Park Street, 2011). Alcoholic

beverages fall into three main categories, beer (4-10% ABV), wine (9-14% ABV) and distilled spirits

(20-95% ABV).

Gaining a better understanding of ethanol’s effect on flavor could aid in production of better

alcohol-free wine and beers, correcting flavor imbalances that occur as we try to continuously

increase the alcohol content of beer and wine, and to gain a better understanding of why many

people prefer distilled spirits in a diluted rather than neat form. In the whiskey industry, master

blenders have been diluting the whiskey to 23% to evaluate the different blends. The most often

noted reason for this is to decrease the pungency of the alcohol, but others also mention that it does

a better job of releasing the flavor profile of the spirit (Smith & Roskrow, 2012). Even though this

been common knowledge in the industry for years, there is no research looking into how those

flavor effects translate to what is happening analytically with the beverage.

28

Additionally, everyone has their favorite way of drinking whiskey. Some people swear that the only

way to drink whiskey is neat, straight out of the bottle at room temperature. Others state that it

needs a splash of water to open up the aroma. Still, others prefer their whiskey on the rock, which

both dilutes and cools the drink. Finally, still others would say that it needs to be diluted closer to

23% ABV to get the best perception of the aroma profile. The proper way to drink whiskey is

obviously personal preference, but there must be obvious flavor differences between these different

drinking styles for people to have a preference and little is known regarding what is driving the

flavor differences.

While it is known that ethanol is an important component to alcoholic beverages, studying these

various drinks pose several problems during analysis. First, the abundance of ethanol makes it

difficult to analyze the other less abundant volatile compounds. The increased concentration of

ethanol can interfere with GC-MS analysis, and it also causes problems for SPME analysis since the

excess ethanol will absorb to the fiber more readily than the other aroma compounds of interest,

making it hard to detect those compounds. Most of the research that has been done utilized static

equilibrium systems, but this is not capable of mimicking what is actually occurring in a beverage

glass during consumption, with the various air currents in the room and the evaporation of ethanol

from the glass. One possible solution is to control the airflow by constructing a shield around the

beverage glass, but this has been shown to create an artificial buildup of odorants over time (Taylor

et al., 2010). Real life dynamic systems are very difficult to monitor but there has been some

improvement in the past decade with the use of real-time mass spectroscopy using, such as

atmospheric pressure chemical ionization (ACPI)-MS that has been modified to use ethanol as the

charge transfer medium (Aznar, Tsachaki, Linforth, Ferreira, & Taylor, 2004).

First, an understanding of the basic physiochemical interactions between water and ethanol is

helpful. The water-ethanol matrix changes significantly as is it changes from being a 100% aqueous

solution to a 100% ethanoic solution. Starting with the aqueous solution, all of the water molecules

are participating in a highly-structured hydrogen-bonded network. This unique property of water is

what gives it such a high surface tension compared to what would be predicted for a molecule of its

size. As ethanol is added to the solution, the ethanol is monodispersed throughout the water until

15%ABV (Conner, Birkmyre, Paterson, & Piggott, 1998; Conner, Paterson, & Piggott, 1999;

D’Angelo, Onori, & Santucci, 1994). At this concentration the ethanol-water matrix changes,

whereby the ethanol molecules in the solution aggregate to form micelles, basically forming a micro

29

emulsion of ethanol in water. This takes place at alcohol concentrations of 17-57% ABV. Once the

ethanol concentration of a solution is above 57%, the remaining water molecules lose their

hydrogen-bonded structure and the solution primarily become ethanoic with the water

monodispersed throughout.

The primary method for monitoring the headspace concentration above alcoholic beverages has

been static headspace analysis. Static headspace analysis is where the beverage is placed in a sealed

vial and then allowed to come to an equilibrium where the volatiles partition into the headspace.

Several studies have looked into the effect of ethanol on the headspace concentration of alcoholic

beverages, and what concentration of ethanol begins to effect the partitioning. Several studies

showed that the critical ethanol concentration is 17%, above which the headspace concentration of

compounds in the ethanol matrix is significantly decreased compared to the 100% water matrix

(Conner et al., 1998; Escalona-Buendia, Piggott, Conner, & Paterson, 1998). Other studies have

shown this effect can be seen at ethanol concentrations as low as 12% (Aznar et al., 2004; Tsachaki,

Aznar, Linforth, & Taylor, 2006). In both situations, the addition of ethanol tends to increase the

solubility of compounds in the matrix thereby decreasing their concentration in the headspace. This

effect is compound specific, with some compounds affected more strongly than others and some

not affected at all by the change in alcohol concentration. The trend seems to follow that the

increase in ethanol makes non-polar compounds more soluble, but this does not hold true for all

compounds.

A newer technique for the analysis of alcoholic beverages has been ACPI-MS which allows for the

evaluation of dynamic systems. When performing the same type of experiment as above except with

continual airflow through the system, the results obtained are reversed. The headspace concentration

above the ethanol solution (12% ABV) is higher than the concentration above the water solution

(Taylor et al., 2010; Tsachaki et al., 2008, 2006; Tsachaki, Linforth, & Taylor, 2005). This

phenomenon can be explained though both the Marangoni effect and the Rayleigh-Bernard

convection, both of which explain the same phenomena on surface tension effects and temperature

density, respectively. The Marangoni effect occurs due to the fact that ethanol lowers the surface

tension of the solution. As a result, it is easier for the ethanol to evaporate, which as it does so

creates areas of higher surface tension due to the ethanol leaving. Since the solution will want to

come to equilibrium, the other ethanol molecules will rearrange, moving to the surface to reduce the

surface tension. In doing this flavor molecules will also move will the ethanol molecules, becoming

30

more available for evaporation. This movement creates a current that essentially stirs the solution,

continually bringing more aroma compounds to the surface for evaporation.

The first study to look at the effects of ethanol on the flavor perception of alcoholic beverages was

done in 1972. In the study entitled “Flavor effects of ethanol on alcoholic beverages,” five

beverages, (cider, wine, sparkling wine, sherry, and whiskey) were evaluated for how their flavor

profile changed when ethanol was removed from the beverages (Williams, 1972). What they found

upon sensory analysis was that when the ethanol was removed, the resulting solution had a stronger

fruity aroma for the cider, wine, and sparkling wine. Likewise, the wine also had increased oxidized

aromas and sour taste compared to the original with ethanol. The sherry was also perceived with

increased sweetness and amyl and hexyl notes with the removal of ethanol. The de-ethanolized

whiskey was perceived as being the most similar to the original compared with the other four

products, but that the removal of ethanol decreased the bite and made the de-ethanolized beverage

drier. Williams and Rosser (1981) then went on to study the effect of ethanol on fruitiness

perception, specifically in cider. In their study, a prepared cider extract was diluted it to volume with

either straight water or various ethanol solutions. Using a sensory panel to compare the aqueous vs.

ethanolic solutions, they found that between 0.1-0.75% ethanol that alcohol increased the fruitiness

perception. Outside of that range, the aqueous solution was always perceived as fruitier.

There has been a significant amount of work looking into the effect of ethanol on whisky, done

primarily by the Scotch Whisky Research Institute. One of their initial studies examined the effect of

ethanol concentration on the solubility of ethyl esters (Conner, Paterson, & Piggott, 1994). Their

results showed that when whiskey is diluted from 40% ABV to 23% ABV that the solubility of the

ethyl esters decreases, creating a super-saturated solutions with esters. As a result, the esters come

together to form agglomerates or micelles, which prevent aroma compounds from partitioning into

the headspace. At the same time, other aldehydes or alcohols may become part of the micelles, or

trapped within it, also affecting their release from the solution. A follow-up study to this examined

the effect of alcohol concentration on the headspace concentration of various ethyl esters (Conner

et al., 1998). Headspace concentrations were calculated for each ethyl ester at alcohol concentrations

of 5%, 10%, 17%, 23% and 40%. At the three lower ethanol levels, the headspace concentrations

were relatively similar, once the 23% solution was reached there was a decrease in concentration by

almost half, and there was another significant decrease at 40%. These headspace concentrations

were converted to activity coefficients (concentration of the compound above the solution over the

31

concentration of the compound above pure compound) and found that for all the esters studies

there was a linear decrease in activity coefficient between 17% and 80% ethanol.

Other studies have also shown the effect that wood extractives can have on the solubility of aroma

compounds. Results demonstrated that not only do wood extractives decrease the size of ester

agglomerates that form (Conner et al., 1994) but that it also decreases the alcohol concentration at

which those micelles will start to form (Conner et al., 1999).

Another study examined the effect of ethanol on headspace concentration of more than just esters.

Authors Boothroyd, Linforth and Cook (2012) examined the effect of ethanol at three different

concentrations (5%, 23% and 40% ABV) of 14 different compounds. Their study showed what the

effect of ethanol on headspace concentration is compound dependent. Several compounds such as

pyrazine, 2-methylpyrazine, 2-acetylthiazole, and furfural were minimally affected by changes in

ethanol concentrations. Other compounds such as β-damascenone and ethyl octanoate were highly

affected by ethanol concentration. While no sensory data was collected to supplement this data, it

can be seen from the analytical data that at each ethanol concentration there is a different ratio of

aroma compounds, this suggests that there will be a different aroma perception at each of these

concentrations. Further work to link the analytical data with what a consumer or panelist perceives is

necessary.

While there has been some work into how ethanol concentration effects flavor perception there is

still much work that needs to be done. First, a better understanding of the dynamic system of flavor

release is required. Continued work using APCI-MS for different ethanol concentrations should be

done to see if the same phenomena observed at 12% alcohol are repeatable at 23% and 40% ABV.

Additionally, there is a great need to be able to link the analytical data collected with sensory data.

32

2.8 Figures

Figure 2.1 Traditional single pot distillation for rum production (with permission Nicol, 2003)

33

Figure 2.2 Diagram of the solera aging system for rum

34

2.9 References

Acree, T. E., & Barnard, J. (1994). Gas chromatography-olfactometry and CharmAnalysis. Trends in

Flavour Research, 35, 211–220.

Arroyo, R. (1939). El Aroma del Ron. Revista de Agricultura, Indistria Y Comercio de Puerto Rico, 31, 448–

450.

Ayala, L. K. (2001). The Rum Experience. Texas, USA: Rum Runner Press Inc.




Barty-King, H., & Massel, A. (1983). Rum: Yesterday and Today. London: Heidelberg Publishers Ltd.



Puerto Rico, 62(4), 330–343.

Blue, A. D. (2004). Rum. In The Complete Book of Spirits: A Guide to Their History, Production, and

Enjoyment (1st ed., pp. 69–103). New York: Harer Collins Publishers.

Bober, A., & Haddaway, L. W. (1963). Gas Chromatographic Identification of Alcoholic Beverages.





Broom, D. (2016). Rum - The Manual. London: Mitchell Beazley.




35

Business Monitor International Research. (2016). United States - Food & Drink Report. Retrieved from

http://search.ebscohost.com/login.aspx?direct=true&db=bth&AN=113556469&site=eds-live

Capone, S., Tufariello, M., & Siciliano, P. (2013). Analytical characterisation of Negroamaro red

wines by Aroma Wheels’’. Food Chemistry, 141, 2906–2915.


Civille, G. V., Lapsley, K., Huang, G., Yada, S., & Seltsam, J. (2010). Development of an Almond

Lexicon To Assess the Sensory Properties of Almond Varieties. Journal of Sensory Studies, 25,

146–162. http://doi.org/10.1111/j.1745-459X.2009.00261.x

Conner, J. M., Birkmyre, L., Paterson, A., & Piggott, J. R. (1998). Headspace Concentrations of

Ethyl Esters at Different Alcoholic Strengths. Journal of the Science of Food and Agriculture, 77,

121–126.

Conner, J. M., Paterson, A., & Piggott, J. R. (1994). Agglomeration of Ethyl Esters in Model Spirit

Solutions and Malt Whskies. Journal of Science of Food and Agriculture, 66, 45–53.

Conner, J. M., Paterson, A., & Piggott, J. R. (1999). Release of distillate flavour compounds in

Scotch malt whisky. Journal of the Science of Food and Agriculture, 79, 1015–1020.

D’Angelo, M., Onori, G., & Santucci, A. (1994). Self-Association of Monohydric Alcohols in Water:

Compressibility and Infrared Absorption Measurements. The Journal of Chemical Physics, 100(4),

3107–3113. http://doi.org/10.1063/1.466452



488. http://doi.org/10.1021/jf0511190

Delavante, M. P. (2004). Rum - The Commercial and Technical Aspects. In J. H. Bryce & G. G.

Stewart (Eds.), Distilled Spirits: Tradition and innovation (pp. 209–214). Nottingham, UK:

Nottingham University Press.

Drake, M. a, McInvale, S. C., Gerard, P. D., Cadwallader, K. R., & Civille, G. V. (2001).

Development of a Descriptive Language for Cheddar Cheese. Journal of Food Science, 66(9),

1422–1427.

36

Eberle, S. (2014, June). Rum’s Handcrafted Renaissance. Outside. Retrieved from

https://www.outsideonline.com/1924131/rums-handcrafted-renaissance

Escalona-Buendia, H., Piggott, J. R., Conner, J. M., & Paterson, A. (1998). Effect of ethanol strength

on the release of higher alcohols and aldehydes in model solutions. Developments in Food Science,

40(C), 615–620.

Fahrasmane, L. (2014). Rum. In C. W. Bamforthm & R. E. Ward (Eds.), The Oxford Handbook of Food

Fermentations (pp. 277–303). New York: Oxford University Press.

Foss, R. (2012). Rum: A Global History. London: Reaktion Books.







Freedman, B. (2016, May). Is High-End Rum the New Whiskey? Forbes. Retrieved from

https://www.forbes.com/sites/brianfreedman/2016/05/19/is-high-end-rum-the-new-

whiskey/#45b3ad364c7f

Gately, I. (2008). Drink: A Cultural History of Alcohol. New York: Gotham Books.

Global Rum Insights - Market Forecasts, Product Innovation and Consumer Trends. (2016).

Retrieved January 1, 2017, from http://www.just-drinks.com/market-research/global-rum-

insights-market-forecasts-product-innovation-and-consumer-

trends_id260118.aspx?utm_source=pr_newswire&utm_medium=press_release&utm_campaig

n=none&utm_term=just_drink/iwsr&utm_content=rum

Global Rum Market: Leading Brands Based on Sales Volume 2015. (2017). Retrieved January 1,

2017, from https://www.statista.com/statistics/259739/leading-rum-brands-worldwide-based-

on-sales-volume/

37

Global Rum Market Set to Decline Over the Next Five Years. (2017, February 14). PR Newswire

Europe. London. Retrieved from http://www.marketwatch.com/story/global-rum-market-set-

to-decline-over-the-next-five-years-2017-02-14-3203148

Global Spirits. (2016). MarketLine Industry Profile.



Gonzáles Gordon, M. M. (1990). The Solera System. In J. Doxat (Ed.), Sherry: The Noble Wine (pp.

141–153). London: Quiller Press.

Hopkins, A. (2014). World’s Largest Rum Markets. Retrieved January 1, 2017, from

https://www.thespiritsbusiness.com/2014/03/worlds-largest-rum-markets/5/

Hopkins, A. (2015). Super-Premium Rums Set for Growth. Retrieved January 1, 2017, from

https://www.thespiritsbusiness.com/2015/11/super-premium-rum-set-for-50-growth/

Irani, D. (2013, November). Renaissance! Rum is Not Just the Favourite Drink of Pirates. The

Economic Times. Retrieved from http://economictimes.indiatimes.com/magazines/brand-

equity/renaissance-rum-is-not-just-the-favourite-drink-of-pirates/articleshow/26421641.cms

Jeffers, H. P. (1997). High Spirits. New York: Lyons & Burford, Publishers.

Jolly, N. P., & Hattingh, S. (2001). A Brandy Aroma Wheel for South African Brandy. South African

Journal of Enology and Viticulture, 22(1), 16–21.

Kleinert, M., Bongartz, A., Raemy, E., & Wadenswil. (2009). Development of a Flavour Wheel for

Breads. Getreide, Mehl Und Brot, 63(1), 60–69.

Koch, I. S., Muller, M., Joubert, E., van der Rijst, M., & Næs, T. (2012). Sensory characterization of

rooibos tea and the development of a rooibos sensory wheel and lexicon. Food Research

International, 46, 217–228. http://doi.org/10.1016/j.foodres.2011.11.028

Labeling and advertising of distilled spirits, Title 27 Code of Federal Regulations, Pt. 5 Section 5:22.

1969 ed.

38

Lawless, L. J. R., Hottenstein, A., & Ellingsworth, J. (2012). The McCormick Spice Wheel: A

Systematic and Visual Approach to Sensory Lexicon Development. Journal of Sensory Studies, 27,

37–47. http://doi.org/10.1111/j.1745-459X.2011.00365.x

Lee, K.-Y. M., Paterson, A., Piggott, J. R., & Richardson, G. D. (2001). Origins of Flavour in

Whiskies and a Revised Flavour Wheel: a Review. Journal of the Institute of Brewing, 107(5), 287–

313. http://doi.org/10.1002/j.2050-0416.2001.tb00099.x

Leppänen, O., Denslow, J., & Ronkainen, P. (1979). A Gas Chromatographic Method for the

Accurate Determination of Low Concentration of Volatile Sulfur Compounds in Alcoholic

Beverages. Journal of the Institute of Brewing, 85, 350–353.




Lurton, L., Ferrari, G., & Snakkers, G. (2012). Cognac: production and aromatic characteristics. In J.

Piggott (Ed.), Alcoholic Beverages: Sensory Evaluation and Consumer Research (pp. 242–266). Oxford:

Woodhead Publishing Limited. http://doi.org/10.1016/B978-0-85709-051-5.50011-0




Macinnis, P. (2002). Bittersweet - The Story of Sugar. Crows Nest, Australia: Allen & Unwin.



Mariani, J. F. (1983). The Dictionary of American Food & Drink. New Haven & New York: Ticknor &

Fields.

Miles, E. (2015). Rum. In Craft Spirit World - A guide to artisan spirit-makers and distillers you need to try

(pp. 70–79). New York: Dog “n” Bone Books.

Minnick, F. (2017). Rum Curious: The Indispensable Tasting Guide to the World’s Spirit. Minneapolis, MN:

39

Voyageur Press.

Mojet, J., & de Jong, S. (1994). The sensory wheel of virgin olive oil. Grasas Y Aceites, 45(1–2), 42–

47.

Monsalve, J. O., López, C., & Zapata, J. (2016). Characterization of aroma compounds in colombian

rums by HS-SPME-GC-MS-O. Revista Colombiana de Química, 45(2), 48–54.




Nicol, D. A. (2003). Rum. In A. G. H. Lea & J. R. Piggott (Eds.), Fermented Beverage Production

(Second, pp. 263–287). New York: Kluwer Academic/Plenum Publishers.

Padgett, T. (2014, April 25). Rum Renaissance Revives The Spirit’s Rough Reputation. NPR- the Salt.

Retrieved from http://www.npr.org/sections/thesalt/2014/04/25/306826984/rum-

renaissance-revives-the-spirits-rough-reputation

Padgett, T. (2017, April). Rum’s Renaissance: Why Rum is having its Cognac Moment. American

Magazine. Retrieved from https://www.americamagazine.org/politics-

society/2017/04/18/rums-renaissance-why-rum-having-its-cognac-moment

Pérez-Cacho, P. R., Galán-Soldevilla, H., Mahattanatawee, K., Elston, A., & Rouseff, R. L. (2008).

Sensory Lexicon for Fresh Squeezed and Processed Orange Juices. Food Science and Technology

International, 14(Suppl. 5), 131–141. http://doi.org/10.1177/1082013208094723




Pino, J. A., & Roncal, E. (2016). Validation of a HS-SPME-GC Method for Determining Higher

Fatty Esters and Oak Lactones in White Rums. Food Analytical Methods, 9, 1958–1962.

http://doi.org/10.1007/s12161-015-0368-3


40





954, 51–57.

Reader, H. P., & Dominguez, M. (2003). Fortified Wines - Sherry, Port and Madeira. In A. G. H.

Lea & J. R. Piggott (Eds.), Fermented Beverage Production (Second, pp. 157–194). New York:

Kluwer Academic/Plenum Publishers.

Rogers, A. (2014). Proof - The Science of Booze. New York: Houghton Mifflin Harcourt.

Schmelzle, A. (2009). The Beer Aroma Wheel. Brewing Science, 62, 26–32. Retrieved from

http://www.scribd.com/doc/201188460/Aroma-Wheel#scribd

Shoup, M. E. (2017). Could rum be the dark horse of spirits? Rum author and importer sees it chip

away at whiskey’s supremacy. Retrieved May 22, 2017, from

http://www.beveragedaily.com/Markets/Rum-might-chip-away-Whiskey-s-supremacy-due-to-

price-increase?utm_source=newsletter_daily&utm_medium=email&utm_campaign=11-May-

2017&c=gZqft5FYEog%2FwSvsxu8NFQ%3D%3D&p2=

Smith, F. H. (2005). Caribbean Rum: A Social and Economic History. Gainsville, FL: University of

Florida Press.

Smith, G. D., & Roskrow, D. (2012). Whiskeyopus. New York: DK Publishing.

Spirits in the United States. (2016). MarketLine Industry Profile.

Stephenson, T. (2017). The Curious Bartender’s Rum Revolution. Ryland Peters & Small.

Stevens, R., & Martin, G. (1965). Gas Chromatographic Identification of Ethyl Esters of Fatty Acids

in Domestic and Imported Rums. Journal of the Association of Official Agricultural Chemists, 48(4),

802–805.

Street, P. (2011). US Alcoholic Beverage Market - Overview. Retrieved November 8, 2015, from

41

http://www.parkstreet.com/industry/alcoholic-beverage-market-overview/

Swartz, K. (2016, February). 8 Trends The Drove the Distilled Spirits Market in 2015. Retrieved

January 1, 2017, from http://www.beveragedynamics.com/2016/02/02/8-trends-that-drove-

the-distilled-spirits-market-in-2015/



2010-1036.ch012







4501(06)80104-4








Ward, C. (2015, May). Rum’s Time. Market Watch. Retrieved from

http://marketwatchmag.com/rum-may-2015/

Williams, A. A. (1972). Flavour Effects of Ethanol in Alcoholic Beverages. The Flavour Industry, 3(12),

604–607.

Williams, A. A., & Rosser, P. R. (1981). Aroma enhancing effects of ethanol. Chemical Senses, 6(2),

42

149–153. http://doi.org/10.1093/chemse/6.2.149

Williams, I. (2005). Rum: A Social and Sociable History of the Real Spirit of 1776. New York: Nation

Books.



43

Chapter 3: Identification of Odor-Active Compounds in a Selection

of Premium Rums

3.1 Abstract

Rum is one of the most diverse distilled spirits on the market. Despite this great product variability,

previous research into the odor-active constituents of rum has only focused on only one or two

rums as representative of the entire class. Therefore, the present study evaluated nine rums, seven

premium and two mixing, to gain a better understanding of overall rum flavor, particularly in terms

better defining the flavor chemistry of premium rums. Fifty-nine odor-active regions, encompassing

sixty-four compounds were identified in the nine rum samples. Aroma extract dilution analysis was

performed on all samples to gain a better understanding of the key potent odorants in rum. The

overall most potent odorants contributing to rum flavor include acetal, 2-/3-methyl-1-butanol, β-

damascenone, 2-phenethyl alcohol, cis-whiskey lactone/4-methylguaiacol, eugenol, sotolon, syringol,

isoeugenol, vanillin, ethyl vanillate, and syringaldehyde. Mixing rums were found to contain the same

odor-active compounds as premium rums but with less potency. Meanwhile, premium rums

contained additional odor-active compounds not present in the mixing rums. Additionally, the

profile of compounds encompassing rum aroma varied from sample to sample.

3.2 Introduction

Rum is a distilled spirit produced from any sugar cane by-product, typically molasses. Rum has a

limited standard of identity, allowing for more variation within the product category than most other

traditional spirits. Rum manufacturers are free to produce rum using whatever starting material (cane

juice, cane syrup or molasses), fermentation style (wild fermentation or culture started), distillation

methods (pot or column distillation), barrels used for maturation (new oak, whiskey, sherry, wine,

etc.), and maturation length they find appropriate.

The earliest studies of rum aroma focused on the identification of all volatile compounds (Batiz &

Rosado, 1978; Bober & Haddaway, 1963; Liebich, Koeing, & Bayer, 1970; Maarse & tem Noever de

44

Brauw, 1966; Ng, 1999; Nykanen, Puputti, & Suomalainen, 1968; Pino et al., 2002; Pino, 2007;

Stevens & Martin, 1965; ter Heide, Schaap, Wobben, de Valosis, & Timmer, 1981; Timmer, ter

Heide, Wobben, & de Valois, 1971; Wobben, Timmer, ter Heide, & de Valois, 1971). While these

studies provided insight into the compounds that are present in rums, they gave no indication about

which volatile compounds contribute to its odor. De Souza and others (2006) were the first to

identify odor-active constituents in rum. Twenty odor-active compounds were identified with β-

damascenone, acetal, ethyl 2-methylbutyrate, and ethyl 3-methylbutyrate being the most potent

odorants. However, no quantitative data was provided at that time. Pino was next to identify

odorants in rums and was the first to quantitate the compounds present (J. A. Pino, Tolle, Gök, &

Winterhalter, 2012). He identified nineteen odor-active compounds and found ethyl butyrate, ethyl

hexanoate, β-damascenone, cis-whiskey lactone, and vanillin to be the most potent odorants in the

rum aroma. Franitza and colleagues (2016a) later identified forty odor-active compounds in two

commercial rums, with the most potent odorants in rum A as cis-whiskey lactone, vanillin, decanoic

acid, 2-/3-methyl-1-butanol, eugenol and sotolon, while rum B was characterized by ethyl

cyclohexanoate, ethyl butanoate, acetal, ethyl 2-methylbutyrate and decanoic acid. Another recent

study evaluated the odorants in nine Colombian rums using HS-GC-MS-O, identifying 46 odor-

active compounds (Monsalve, Lopez, & Zapata, 2016). No quantitation of the identified volatiles

was performed. Franitza also went on to study the production process of rum following 44

compounds and how their concentrations increased or decreased during fermentation, distillation,

and maturation (Franitza, Granvogl, & Schieberle, 2016b).

While a significant amount of research has been done on rum aroma, the previous studies have only

focused on one or two rum samples, except the Colombian rum study. Due to the considerable

variation in production styles of rum, it is difficult to say that the previously evaluated rums

encompass the totality of rum as a class. The goal of this study was to identify the odor-active

compounds in a variety of premium aged rums and several mixing rums, in order to gain a better

understanding of the important odorants in premium aged rums and how they are differentiated

from mixing rums.

45

3.3 Materials and Methods

Materials

Nine rums, seven premium and two mixing rums, were chosen for analysis based on expert ratings,

awards received and product availability (Table 3.1). The nine selected rums were purchased at a

local liquor store (Champaign, IL). All nine rums had reported ethanol concentration of 40% alcohol

by volume (ABV). Mention of the brand name of these rums does not imply any research contact or

sponsorship and is not for advertisement or endorsement purposes.

De-odorized water used for volatile extraction was prepared by boiling deionized-distilled water to

two-thirds of its original volume.

Chemicals

Dichloromethane and anhydrous sodium sulfate were purchased from Fisher Scientific Co. (Fair

Lawn, NJ) and used for volatile extraction.

Reference Standard Compounds

The following chemicals were used as authentic standards to confirm the identification of odor

compounds: acetaldehyde and isobutanol and were obtained from Fisher (Fair Lawn, NJ); 2-

methylpropanal, ethyl propanoate, 2,3-butanedione, ethyl-3-methylbutyrate, hexanal, isoamyl acetate,

ethyl hexanoate, ethyl octanoate, methional, phenyl acetaldehyde, p-cresol, eugenol, and Z-6-

dodecene-γ-lactone were obtained from Aldrich (Milwaukee, WI); 3-methylbutanal, ethyl 2-

methylpropanoate, ethyl butyrate, ethyl-2-methylbutyrate, ethyl pentanoate, heptanal, 2-methyl-1-

butanol, 3-methyl-1-butanol, 1-octen-3-one, (E)-2-nonenal, 3-methylbutyric acid, mixture of cis-&-

trans whiskey lactone, 2-phenethyl alcohol, guaiacol, 4-propyl guaiacol, syringol, and ethyl vanillin

were obtained from Sigma-Aldrich (St. Louis, MO); acetic acid, ethyl 3-phenylpropanoate, 4-methyl

guaiacol, 4-ethyl guaiacol, m-cresol, sotolon, vanillin, syringaldehyde were obtained from SAFC (St.

Louis, MO); isoeugenol and ethyl vanillate were obtained from Alfa Aeasar (Lancaster, UK); 2-

methylbutanal and 1-octen-3-ol were obtained from Bedoukian (Danbury CT); acetal (1,1-

diethoxyethane) was obtained from Acros Organics (NJ); β-damascenone was obtained from

Firmenich (Switzerland); ethyl acetate was obtained from Applied Biosystems Inc. (Foster City, CA).

46

Ethyl cyclohexancarboxylate was synthesized as follows below.

Synthesis of Ethyl Cyclohecancarboxylate

Cyclohexanecarboxylic acid (1.28 g; 10 mmol), ethanol (4.6 g, 100 mmol) and 2 drops of

concentrated H2SO4 were added to a 22-mL glass vial. The vial was sealed with a PTFE-lined silicon

cap and then heated at 100C for 2 h. After cooling the vial to room temperature the reaction

mixture was diluted 10 mL of pentane and extracted with a saturated aqueous Na2CO3 solution (2 x

5 mL). The pentane layer was washed with saturated aqueous NaCl (2 x 5 mL) and dried over

anhydrous Na2SO4. The target compound (0.91 g) was recovered after removal of the pentane

using a gentle stream of N2 gas.

Methods

Direct Solvent Extraction

Nine rums, seven premium and two mixing rums, were pipetted (10mL) into individual 50mL

screwcap centrifuge tubes. To reduce the alcohol by volume ratio to 10% ethanol, 30mL of

deodorized water was added. The tubes were sealed with PTFE-lined caps and shaken for 5 minutes

by hand. Dichloromethane (2mL) was added and the mixture shaken for an additional 5 minutes.

The tubes were then centrifuged at 3500 RMP for 5 minutes (IEC HN-SI; Damon/IEC Division,

Needham Heights, Massachusetts) to separate the solvent layers. The bottom phase

(dichloromethane) was transferred into a 15mL conical test tube. The extraction with

dichloromethane was repeated two more times, and the extracts were pooled. The pooled extract

was dried with sodium sulfate (2g) to remove any excess water. The final dried extract was

condensed to 0.5 mL using a gentle stream of nitrogen gas and stored at -20° C prior to analysis.

GCO Parameters

The gas chromatography-olfactometry (GCO) system used for analysis of the rum extracts consisted

of a 6890N GC (Agilent Technologies, Inc., Wilmington, DE) equipped with a FID and olfactory

detection port (OD2, Gerstel, Germany). One μL of extract was injected into a CIS-4 inlet (Gerstel,

Germany) in the cold splitless mode (-50° C for 0.10 min, then increased at 12° C/sec to a final

temperature of 260° C). Polar separation was performed using a RTX®-Wax column (15 m length x

0.53 mm i.d. x 1.0 μm film thickness; Restek; Bellefonte, PA), while non-polar separation was

47

performed using RTX®-5MS column (15 m length x 0.53 mm i.d. x 1.0 μm film thickness; Restek)

to aid in identification. Helium was used as the carrier gas at 5.0 mL/minute. Post-separation, the

flow was split using a deactivated, uncoated fused silica gel column between the FID (250° C) and

the olfactory port (transfer line 250° C). Oven temperature was programmed as follows: initial

temperature, 40° C (5 min hold), ramp rate 10° C/min, final temperature, 225° C (30 min hold).

GC-MS Parameters

The GC-MS system used for analysis of rum extracts consisted of a 6890N GC equipped with a

5973N mass spectrometer (Agilent Technologies Inc.). One μL of spiked extract was injected into a

CIS-4 inlet (Gerstel, Germany) in the cold splitless mode (-50° C for 0.10 min, then increased at 12°

C/sec to a final temperature of 260° C). Separations were performed using a RTX®-Wax column

(30 m length x 0.53 mm i.d. x 1.0 μm film thickness; Restek; Bellefonte, PA). Helium was used as

the carrier gas at 5.0 mL/minute. FID temperature was 250° C. Oven temperature was programmed

as follows: initial temperature, 40° C (5 min hold), ramp rate 4° C/min, final temperature, 225° C

(30 min hold). To aid in identification, analysis was also conducted using a RTX®-5MS column (30

m length x 0.53 mm i.d. x 1.0 μm film thickness; Restek).

Identification of Compounds

The rum extracts were subject to evaluation by both GCO and GC-MS. The retention index (RI)

was calculated for each aroma active compound based on comparing its retention time (RT) to those

of standard n-alkanes (van Den Dool & Kratz, 1963). The RI is calculated using the equation:

𝑅𝐼 = 100𝑛 + 100𝑛𝑑𝑖𝑓𝑓 [(𝑡𝑟(𝑢𝑛𝑘𝑛𝑜𝑤𝑛) − 𝑡𝑟(𝑛))

(𝑡𝑟(𝑁) − 𝑡𝑟(𝑛))]

where the difference in retention time between the unknown (𝑡𝑟(𝑢𝑛𝑘𝑛𝑜𝑤𝑛)) and the lower alkane

(𝑡𝑟(𝑛)), is divided by the difference between the retention time of the upper alkane (𝑡𝑟(𝑁)) and the

lower alkane, then multiplied by 100 times the difference in carbon numbers of the two alkanes

(𝑛𝑑𝑖𝑓𝑓), and added to 100 times the number of carbons in the lower alkane(𝑛). Odor-active

compounds were tentatively identified based on three criteria: 1) comparison of RI on two different

stationary phase columns (RTX-5 and wax) against literature values for known compounds, 2)

comparison of the odor properties against published values, and 3) comparison of electron

48

ionization mass spectra (EI-MS) from the GC-MS to the EI-MS spectra in the National Institute of

Standards and Technology (NIST) database. Compounds were considered positively identified when

these comparisons were made against data from authentic reference standard analyzed under

identical conditions.

Aroma Extract Dilution Analysis

Starting with a 0.5 mL aroma extract, AEDA was performed on a stepwise dilution series in

dichloromethane. For this, 100 µL of the 0.5mL concentrated extract was diluted into 100 µL of

dichloromethane serially to obtain 1:2 (FD=2), 1:4 (FD=4), 1:8 (FD=8), 1:16 (FD=16), 1:32

(FD=32), 1:64 (FD=64), 1:128 (FD=128), 1:256 (FD=256), 1:512 (FD=512), 1:1024 (FD=1024),

1:2048 (FD=2048), 1:4096 (FD=4095), and 1:8192 (FD=8192), dilution ratios. Each dilution was

stored in a 1.5 mL septum-capped Target DP vial (National Scientific, Rockwood, TN) at -20° C

prior to analysis. Dilution series were evaluated in order, starting with the most concentrated sample

and subsequent dilutions were evaluated until no odorants were detected. The last dilution a

compound was perceived was noted, corresponding to the compounds dilution factor. Due to time

constraints, multiple panelists performed AEDA. The final flavor dilution (FD) factors were

determined by converting individual FD factors to log2 values, averaging the results from two

panelists rounding up to the nearest whole number and finally converting the values back to FD

factors.

3.4 Results and Discussion

Nine rum samples, consisting of two mixing rums and seven premium rums were chosen for

comprehensive aroma analysis (Table 3.1). Bacardi Superior and Bacardi Gold were chosen to

represent the category of mixing rums since Bacardi is the number one seller of rum in the United

States and the number two rum brand worldwide after McDowell’s No. 1 Celebration (an Indian

rum brand) (“Leading rum brands worldwide in 2015, based on sales volume (in million 9 liter

cases),” 2017). The seven premium rums were chosen by comparing the review score on the Rum

Howler blog (an industry expert), the number of awards and medals received by the company,

making sure that a variety of different rums (in terms of production style) were evaluated and that

the rum was able to be purchased at a local store.

49

Identification of Odorants and Aroma Extract Dilution Analysis

All rum samples were extracted and analyzed by GCO on two separate columns to identify the

odor-active compounds present in the nine rum samples. Identification of the odor-active regions

was determined by comparison of the retention index (RI) on two columns (RTX-Wax and RTX-5)

to literature values, odor property matching that of literature, comparison of mass spectrum of the

compounds determined by EI-MS to a library of known compounds and verification with an

authentic standard. AEDA was performed on all rum samples to determine the relative potency of

the identified compounds. FD factors are calculated on the basis of the compound’s odor activity

value (OAV) in air. Compounds perceived at higher dilutions are assumed to have a greater impact

on the overall aroma of rum. It is important to note that OAV’s for compounds are highly

dependent on the matrix and while the compound may be perceived easily in air, it may be below its

OAV is an alcoholic matrix.

Most Potent Odorants in Individual Rums

The intensity and order of potent odorants was found to vary between rum samples, as was the

number of potent odorants detected. The total number of odor regions perceived for each rum and

the most potent odorants (FD≥32) are reported below for each rum. The complete list of

compounds identified and corresponding FD factors for all rums can be found in Table 3.2.

Odor-Active Compounds in Bacardi White Rum

Nineteen odor-active regions encompassing 22 compounds were detected in the BW rum. The

region with the highest FD factor was identified as vanillin (FD=128), followed by acetal and 2-/3-

methyl-1-butanol (FD=32). All other regions had FD factors of 8 or less.

Odor-Active Compounds in Bacardi Gold Rum

Thirty-three odor-active regions were detected in the BG rum. The regions with the highest FD

factors were found to be 2-phenethyl alcohol, (E)-isoeugenol, and vanillin (FD=128). The next

highest regions were 2-/3-methylbutanal and β-damascenone (FD=64) followed by cis-whiskey

lactone/4-methylguaiacol, and sotolon (FD=32).

50

Odor-Active Compounds in Appleton Estate Rum

Forty-three odor-active regions were detected in AE rum. The regions with the highest FD factor

were 2-/3-methyl-1-butanol, and 2-phenethyl alcohol (FD=512). The next most potent regions were

cis-whiskey lactone/4-methylguaiacol (FD=256), acetal, sotolon, and vanillin (128), followed by (E)-

isoeugenol and syringaldehyde (FD=64).

Odor-Active Compounds in Ron Abuelo 7 year Rum

Forty-four odor-active regions were detected in RA7. The region with the highest FD factor was

vanillin (1024). This was followed by 2-phenethyl alcohol (FD=256), and 2-/3-methyl-1-butanol, cis-

whiskey lactone/4-methylguaiacol (FD=128). Medium potent odorants consisted of acetal, sotolon,

syringol, and ethyl vanillate (FD=64) and syringaldehyde (FD=32).

Odor-Active Compounds in Appleton Estate 12 year Rum

Forty-five odor-active regions were detected in AE12. The two most potent regions were

determined to be vanillin (FD=2048) and (E)-isoeugenol (FD=512). The next most potent regions

were 2-phenethyl alcohol (FD=128), followed by eugenol, ethyl vanillate, and unknown 59

(FD=64). Acetal, 2-/3-methyl-1-butanol, cis-whiskey lactone/4-methylguaiacol, and sotolon regions

were also found to have a medium odor potency (FD=32).

Odor-Active Compounds in El Dorado 12 Year Rum

Fifty odor-active regions were detected in ED12. The most potent odorant was found to be vanillin

(FD=512), followed by cis-whiskey lactone/4-methylguaiacol (FD=256) 2-phenethyl alcohol, and

syringol (FD=128). Medium potent odor regions were determined to be 2-/3-methyl-1-butanol,

trans-whiskey lactone/ethyl 3-phenylpropanoate (FD=64), sotolon, and (E)-isoeugenol (FD=32).

Odor-Active Compounds in Diplomatico Reserva 12 year Rum

Forty-four odor-active regions were detected in DR12 rum. The most potent odorants were

determined to be vanillin (FD=4096), followed by ethyl vanillin (FD=2048), 2-/3-methyl-1-butanol

(FD=1024), and 2-phenethyl alcohol (FD=512). (E)-Isoeugenol, unknown 59 (FD=256) and acetal

(FD=128) were also found to have high potency. Medium potent odorants were determined to be

51

sotolon and syringaldehyde (FD=64) followed by ethyl butyrate, β-damascenone, trans-whiskey

lactone/ethyl 3-phenylpropanoate, cis-whiskey lactone/4-methylguaiacol and ethyl vanillate

(FD=32).

Odor-Active Compounds in Ron Zacapa Rum

Forty-five odor-active regions were detected in RZ. The most potent odorants were determined to

be vanillin (FD=4096), 2-phenethyl alcohol (FD=2048), 2-/3-methyl-1-butanol, and cis-whiskey

lactone/4-methylguaiacol (FD=1024). Medium potent odorants were determined to be β-

damascenone (FD=64) followed by guaiacol, m-cresol, sotolon, (E)-isoeugenol, syringaldehyde and

unknown 59 (FD=32).

Odor-Active Compounds in Dictador Insolent XO Rum

Thirty-six odor-active regions were detected in RZ. The region with the highest FD factor was ethyl

vanillin (2048) followed by cis-whiskey lactone/4-methylguaiacol (512), vanillin (256), and 2-

phenethyl alcohol (128). Regions with medium potency were 2-/3-methyl-1-butanol (FD=64),

unknown 40, and isoeugenol (FD=32).

Overall, 59 odor-active regions, consisting of 64 compounds were detected in the nine rum samples.

FD factors for all nine rums are shown in Table 3.2. Of those 59 regions, 11 were detected in all

nine rums including acetal, ethyl butyrate, ethyl 2-methylbutyrate, 2-/3-methyl-1-butanol, trans-

whiskey lactone/ethyl 3-phenylpropanoate, 2-phenethyl alcohol, cis-whiskey lactone/4-

methylguaiacol, ethyl vanillin, vanillin, syringaldehyde and unknown 59 (Wax RI-2951, campfire).

Twelve additional regions were detected in all rums except BW consisting of acetaldehyde, 2-/3-

methylbutanal, ethyl 2-methylpropanoate, (Z)-2-nonenal, unknown 32 (Wax RI-1724, fruity, spice),

β-damascenone, eugenol, sotolon, syringol, (E)-isoeugenol, unknown 51 (Wax RI-2421, woody), and

ethyl vanillate. The absence of these compounds in BW suggests that they are formed during the

maturation process, either through extraction or reactions within the distillate, or that these

compounds are lost during the charcoal filtration step used to remove the color from BW rum.

Additionally, these twenty-three regions encompassed all the most potent odorants (FD≥64)

detected in the nine rum samples. Vanillin was found to be the most potent odorant in most rum

samples except AE (2-phenethyl alcohol and 2-/3-methyl-1-butanol) and DX (ethyl vanillin). The

overall most potent odorants contributing to rum flavor include acetal, 2-/3-methyl-1-butanol, β-

52

damascenone, 2-phenethyl alcohol, cis-whiskey lactone/4-methylguaiacol, eugenol, sotolon, syringol,

(E)-isoeugenol, vanillin, ethyl vanillate, and syringaldehyde.

Unknowns accounted for fourteen of the detected odor-active regions. None of the unknowns are

perceived at FD factors higher than 16 in any of the samples with the exception of unknown 59

(Wax RI-2951, Sac-5 RI-1832, campfire). This compound was found in every sample, with DR12

having the highest perception of the odorant (FD=256). Unknown 59 could not be identified

despite having a significant peak. Electron ionization (EI)-MS (Figure 3.1) suggests the compound

has a molecular weight of 226. However, no confirmed matches were found from searching the

NIST database. Further research is required to identify this compound. The relatively high

abundance of this compound suggests it may contribute to woody and smoky perceptions in rum.

Of the 50 odorants identified in the present study, 40 were previously identified as odor-active in

rum (Burnside, 2012; de Souza et al., 2006; Franitza et al., 2016a, 2016b; J. A. Pino et al., 2012). In

total, 10 compounds were identified for the first time as odor-active in rum, specifically:

acetaldehyde, ethyl acetate, 1-octene-3-ol, (Z)-2-nonenal, m-cresol, syringol, (E)-isoeugenol, Z-6-

dodecene-γ-lactone, ethyl vanillin, and syringaldehyde. The majority of these compounds have been

previously identified in rum (Batiz & Rosado, 1978; Liebich et al., 1970; Maga, 1989; Pino, 2007;

Pino, Marbot, Perez, & Nunez de Villavicencio, 1999; Pino et al., 2012; ter Heide et al., 1981).

However, this is the first study to identify (Z)-2-nonen-al, (Z)-6-dodecene-γ-lactone and ethyl

vanillin in any capacity in rum. (Z)-6-Dodecene-γ-lactone has previously been identified in Bourbon

whiskey (Poisson & Schieberle, 2008) and white wine (Guth, 1997) and (Z)-2-nonenal has previously

been detected in wood extracts (de Simón, Esteruelas, Muñoz, Cadahía, & Sanz, 2009). The

presence of ethyl vanillin in the rum samples was not expected. Ethyl vanillin is a synthetic

compound that is typically used as artificial vanilla. Ethyl vanillin was most likely added to the rum

samples to increase vanilla aroma which is permissible by the Alcohol and Tobacco Tax Trade

Bureau (Limited Ingredients, 2016). Under the regulation, ethyl vanillin and vanillin can be added to

distilled beverages as long as the total concentration does not exceed 40ppm of vanillin and ethyl

vanillin, and the rum does not need to be labeled as artificial.

53

Comparison of premium and mixing rums

Mixing rums were found to contain the same odorants as the premium rums with the exception of

ethyl propanoate and 2,3-butandione which were the only two odorants to be detected exclusively in

the mixing rums. All compounds detected in BW were also detected at the same or higher dilution

factors in BG except heptanal and unknown 55 (Wax RI-2578, fresh/floral). The mixing rums

contained less odor-active compounds. These compounds were also present in lower abundance, as

indicated by the lower dilution factors, compared to the premium rum samples. All compounds

detected in the mixing rums, except for the two previously mentioned, are found in at least two

premium rums, with the majority of detected compounds found in all premium rums.

Premium rums were found to have at least ten additional odor-active regions compared to the

mixing rums, with the exception of DX (only 36 odor-active regions). Due to the large-scale of

Bacardi rum production, continuous column distillation is used. This process removes more of the

volatile compounds produced during fermentation than a traditional pot still. Additionally, the

increase in the number of odorants perceived in the premium rums could be a result of the extended

time spent in the casks compared to the mixing rums. The extended time in the cask allows for the

extraction of more oak extractives as well as allowing time for more reactions to occur between

compounds in the distillate.

Premium rums all contained the following 23 compounds: acetaldehyde, acetal, 2-/3-methylbutanal,

ethyl 2-methylpropanoate, ethyl butyrate, ethyl 2-methylbutyrate, 2-/3-methyl butan-1-ol, (Z)-2-

nonenal, unknown 32 (fruity/spicy), β-damascenone, trans-whiskey lactone/ethyl 3-

phenylpropanoate, 2-phenethyl alcohol, cis-whiskey lactone/4-methylguaiacol, eugenol, sotolon,

syringol, (E)-isoeugenol, unknown 51 (woody), ethyl vanillin, vanillin, ethyl vanillate, syringaldehyde,

and unknown 59 (campfire). Additional compounds were perceived in at least two samples with the

exception of (E)-2-nonenal which was only found in ED12. Additionally, the potency and

abundance of compounds varied from rum to rum. These variations are likely attributed to

differences in production style. However, it is difficult to draw conclusions as to how specific

difference in production methodologies may change the volatile profile of rum with the present

information. A more thorough investigation into how production practices alter final rum is needed.

54

Origins of Odorants Found in Rums

Odorants present in the final rums can come from a variety of sources. These include coming from

the initial molasses or cane juice, formation during fermentation, created from reactions occurring

during distillation and/or aging, or extracted from the wood barrels.

The starting material, either molasses, cane juice or cane syrup, contributes more than just

fermentable sugar to the final rum product. Franitza and others (2016b) followed odor-active

compounds during the production process of rum, demonstrating that all odor-active compounds

identified in the final rum product were present in the initial molasses, at some concentration, with

the exception of cis- & trans-whiskey lactone. Other studies have also identified acetic acid,

damascenone, guaiacol, 2-phenethyl alcohol, p-cresol, sotolon, syringol, isoeugenol, and vanillin in

sugar cane molasses (Abe, Nakatani, Yamanishi, & Muraki, 1978a, 1978b; Tokitomo, Kobayashi,

Yamanishi, & Muraki, 1980). Kobatashi (1989a) demonstrated sotolon to be a key odorant in raw

sugar cane aroma.

Ethanol is the main by-product and reason for fermentation; however, numerous other odor-active

compounds are also generated from the yeast and bacteria. The main compounds generated during

fermentation are alcohols, acids, and esters. Saccharomyces cerevisiae and Schizosaccharomyces pombe help to

generate propanol, butanol, isobutanol, propanoic acid, butyric acid, isobutyric acid, 3-methyl

butanoic acid, and hexanoic acid (Fahrasmane, Parfait, Jouret, & Galzy, 1985). Formation of ethyl

acetate, isoamyl acetate, and acetic acid by yeast has also been demonstrated (Nordström, 1963; B. L.

Nykänen, Nykänen, & Suomalainen, 1977; L. Nykänen & Nykänen, 1977). Ethyl esters are formed

enzymatically by activating the corresponding acids to acyl-CoA, which then reacts with ethanol.

(Nordström, 1963). Ethyl 2-methylbutanoate is formed enzymatically from the esterification of 2-

methyl-1-butanol with activated acetic acid (Matheis, Granvogl, & Schieberle, 2016). In alcoholic

beverages, only the (S)-isomer exists as 2-methylbutanol is formed from (S)-isoleucine. 2-Methyl

butan-1-ol is partially formed from the reduction of 2-methylbutanal according to the Ehrlich

mechanism by oxidation of 2-methylbutanol to 2-methylbutanoic acid, followed by esterification

with ethanol. However, this cannot be the only mechanism of formation as both (R) and (S)

enantiomers of 2-methylbutanal have been found in alcoholic beverages. 2-methylbutanal can also

be formed through Strecker-type degradations of L-isoleucine.

55

Aside from being present in the initial molasses used for fermentation, β-damascenone significantly

increases after distillation (Franitza et al., 2016b). The formation of β-damascenone in alcoholic

beverages is still unclear, although it is hypothesized that precursors could be generated during

fermentation and then converted to β-damascenone through acid hydrolysis (Sefton,

Skouroumounis, Elsey, & Taylor, 2011).

Maturation also has a significant impact on the final aroma of rum. Compounds formed during

maturation come from three main pathways: 1)extracted directly from the wood, 2) decomposition

of the wood macromolecules and 3) reactions that occur in the barrel between wood extractive,

distillate components or both (Mosedale, 1995). Compounds directly extracted from the wood

include compounds such as 2-phenethyl alcohol, 4-propylguaiacol, m-cresol, and cis- & trans-whiskey

lactone (Cutzach, Chatonnet, Henry, & Dubourdieu, 1997; Maga, 1989; Mosedale, 1995). Cis- &

trans-whiskey lactones were confirmed as only coming from wood extractives by Franitza and others

(Franitza et al., 2016b), as the whiskey lactones were not detected in any of the steps prior to

maturation.

Charring of the barrels has also been shown to form the precursors for (E)-2-nonenal and 1-octen-

3-one which are then generated through auto-oxidative reactions, which are then extracted by

ethanol into the distillate (Chatonnet & Dubourdieu, 1998). Whiskey lactone has been shown to be

generated during barrel charring (Conner, Paterson, & Piggott, 1993).

The decomposition of wood macromolecules can be initiated by two methods. The first is through

extraction of the lignin into the distillate, a process known as ethanolysis (Baldwin, Black,

Andreasen, & Adams, 1967; Mosedale & Puech, 1998). The lignin is then broken down into sub-

units of either coniferyl alcohol, sinapyl alcohol, and coumaryl alcohol. These lignin sub-units can

also be formed through pyrolysis of the lignin during barrel charring (Conner et al., 1993; Piggott &

Conner, 2003). The sub-units can further react and breakdown to form guaiacol, 4-ethylguaiacol, 4-

vinylguaiacol, eugenol, isoeugenol and vanillin from coniferyl alcohol; syringaldehyde and syringol

from coumaryl alcohol; and p-cresol from sinapyl alcohol (Baldwin et al., 1967; Cutzach et al., 1997;

Genthner, 2014; Maga, 1989).

Chemical reactions that occur, typically oxidation and acetal formation, in the distillate during aging

can account for the formation compounds such as acetaldehyde and acetic acid (Piggott & Conner,

56

2003). Esterification is one of the most frequent reactions that takes place between wood extractive

such as aliphatic acids and compounds originally from the distillate including the formation of ethyl

esters such as ethyl acetate, ethyl octanoate, ethyl hexanoate, and ethyl vanillate due to the significant

amount of ethanol present (Maga, 1989; Piggott & Conner, 2003).

Sotolon was initially demonstrated to be formed from glutamic acid and pyruvate in raw sugar cane

were the glutamic acid oxidized to α-keto glutarate and then condensed with the pyruvate

(Kobayashi, 1989). It was later demonstrated that sotolon was formed from α-ketobutyric acid and

acetaldehyde in wine model systems (Thuy, Elisabeth, Pascal, & Claudine, 1995). Sotolon

concentration has been shown to increase over time in both Madeira wine and port (Câmara, Alves,

& Marques, 2006; Silvia Ferreira, Barbe, & Bertrand, 2003). Other reaction mechanisms have been

suggested sotolon could be formed from the condensation of diacetyl and hydroxyacetaldehyde

(Silvia Ferreira et al., 2003). The most likely pathway in the aldol condensation between 2-

ketobutyric acid and acetaldehyde, as has also demonstrated in dry white wines, and results have

shown that 2-ketobutyric acid can be formed by strains of Saccharomyces cerevisiae (Pons, Lavigne,

Landais, Darriet, & Dubourdieu, 2010).

The effect of wood maturation on the generation of these compounds in rum was demonstrated in

Franitza’s (2016b) recent study following the production process of rum, were the concentrations of

vanillin, guaiacol, 4-ethylguaiacol, 4-propylguaiacol, p-cresol, sotolon and cis- & trans-whiskey lactone

all increased during maturation.

Conclusion

A total of 10 compounds were identified as odor-active compounds for the first time in rum. The

compounds (Z)-2-nonenal, (Z)-6-dodecene-γ-lactone and ethyl vanillin were identified for the first

ever in any rum sample. In summary, clear differences exist between mixing and premium rums,

primarily in number and potency of odorants as well as with the premium rum class. However, the

23 odorants identified in all premium samples suggest they are the key components of rum flavor.

Differences in production practices are the most likely reason for variations between samples.

Quantitative studies to confirm the contribution of the identified compound to the overall aroma of

rum have been undertaken and will be discussed in Chapter 4.

57

3.5 Tables and Figure

Table 3.1 Product and manufacturing information for rums analyzed

Code Rum Age* Country of Origin Barrels Used Manufacturer

BW Bacardi SuperiorM 1 year Puerto Rico White OakF Bacardi Limited

BG Bacardi GoldM 2 years Puerto Rico Toasted OakF Bacardi Limited

AE Appleton Estate V/XM 5-10 yearsB Jamaica Whiskey Barrels J. Wary & Nephew Ltd.

RA7 Ron Abuelo: Reserva SuperiorM

7 years Panama Small Oak Barrels Varela Hermanos

AE12 Appleton Estate ExtraM 12 years Jamaica American Oak BarrelsP J. Wary & Nephew Ltd.

DR12 Diplomatica Reserva Exclusiva C

12 years Venezuela Small Oak CasksP Destilería Unidas S.A

ED12 El Dorado: Finest Demerara RumM

12 years Guyana Bourbon Oak Casks+ Demerara Distillers Limited

RZ Ron Zacapa (Centenario) XO: Solera Gran Reserva

EspecialC

6-25 yearsS Guatemala American Whiskey, Sherry, Pedro Ximenez

Wines and Cognac

Rum Creation and Products,

Inc.

DX Dictador XO InsolentC 21 yearsS Columbia Jerez and Port BarrelsA, R

Dictador

*Age declared on bottle, M Produced from molasses, C Produced from sugar cane honey, B Blend of rums, S aged using Solera system, F Rum charcoal filtered, P Distilled using copper pot stills, + Combination of wooden and metal coffee stills and wooden pot stills, A Distilled in stainless steel alembic, R Barrels recharred

58

Table 3.2 Odor-active compounds identified by AEDA of nine rums

RIa

Flavor Dilution Factorb,c

No. Compound Odor Descriptiond WAX RTX 5

BW BG AE RA7 AE12 DR12 ED12 RZ DX

1 acetaldehydee fruity, sweet 712 <500 -f 1 2 8 2 2 4 1 2

2 2-methylpropanale chocolate 906 592 - - 1 - 2 2 - 1 2

3 ethyl acetate plastic 922 629 - - - - 1 - 1 2 -

4 acetale melon 927 720 32 16 128 64 32 128 16 2 16

5a,b 2-/3-methylbutanale chocolate 930 669 - 2 16 2 4 2 4 8 1

6 ethyl propanoatee cherry 960 711 2 2 - - - - - - -

7 ethyl 2-methylpropanoatee fruity 968 752 - 4 8 8 8 1 4 2 2

8 2,3-butanedioneg sweaty/cheesy 983 598 1 1 - - - - - - -

9 unknown plastic 998 1 2 2 - 4 1 4 - -

10 unknown pungent, plastic, fruity 1010 - 1 - - 4 2 1 - -

11 unknown solvent, painty 1014 - - - 2 2 - 4 4 -

12 ethyl butyratee fruity 1041 803 1 4 2 4 2 32 2 1 8

13 ethyl 2-methylbutyratee melon 1057 844 1 1 16 8 8 1 8 1 2

14 unknown fruity 1061 2 16 4 4 2 2 4 2 -

15 ethyl 3-methylbutyratee fruity 1075 852 1 4 1 1 - - 2 - -

16 hexanalh green 1087 807 - - - - - - 1 - 1

17 isobutanole chocolate 1101 653 - - - 1 1 4 1 2 -

18 isoamyl acetatee fruity 1122 884 - - - - - 1 1 2 -

19 ethyl pentanoatee fruity 1145 900 - - - 1 - 1 1 - 1

20 heptanalg fruity 1197 900 1 - - - - 1 1 - -

21a,b 2-/3-methyl-1-butanole chocolate 1212 731 32 64 512 128 32 1024 64 1024 64

22 ethyl hexanoatee floral 1239 1001 - - - - 1 1 2 1 1

23 1-octen-3-oneh mushroom, earthy 1303 980 1 2 1 1 - 1 1 2 1

24 ethyl cyclohexanecarboxylateg fruity 1422 1126 - - 16 2 - 2 2 2 -

25 ethyl octanoatee earthy, soil 1436 1195 - 2 2 4 - - - - -

26 acetic acide fruity, vinegar 1448 630 - - 1 1 4 2 - 2 2

27 methionalg/1-octene-3-olh potato/mushroom 1460 902/980 - - 1 4 2 2 - 4 -

59

Table 3.2 (cont.) Odor-active compounds identified by AEDA of nine rums

RIa Flavor Dilution Factorb,c

No. Compound Odor Descriptiond WAX RTX 5 BW BG AE RA7 AE12 DR12 ED12 RZ DX

28 (Z)-2-nonenalg laundry, floral 1502 1148 - 1 2 16 8 2 8 4 2

29 (E)-2-nonenalg hay 1536 1141 - - - - - - 2 - -

30 phenyl acetaldehydeh floral 1636 - - - - 2 2 2 - 1

31 3-methylbutyric acidg cheesy 1661 856 - - 8 2 1 - 1 16 2

32 unknown fruity/spicy 1724 - 8 8 4 2 1 4 16 2

33 unknown floral 1798 - - - - 1 - 2 - -

34 β-damascenonee applesauce 1813 1380 - 64 32 16 16 32 8 64 1

35 guaiacole spices/cloves 1844 1084 - 1 8 2 4 - 2 32 -

36a,b trans-whiskey lactonee/ ethyl 3-phenylpropanoateh

floral 1876 1290/1350

8 16 8 8 8 32 64 8 8

37 2-phenethyl alcohole roses 1903 1109 8 128 512 256 128 512 128 2048 128

38a,b cis-whiskey lactonee/ 4-methylguaiacole

coconut, potpourri 1939 1311/1182

2 32 256 128 32 32 256 1024 512

39 unknown spicy, cloves 1961 - - 1 - - - 1 4 -

40 unknown cotton candy 1998 - - 32 2 - 1 4 2 32

41 4-ethylguaiacole spices, floral 2016 1270 - - 2 2 8 4 1 - 4

42 unknown sweat BO 2033 - - 4 - - - 2 2 -

43 p-cresole skunky/barnyard 2070 1076 - - 16 1 2 4 - 1 -

44 m-cresole burnt plastic 2076 1076 - - 16 8 2 8 - 32 1

45 4-propylguaiacolh nutmeg 2095 - - 8 8 4 4 - 4 4 -

46 eugenole baking spices 2152 1367 - 8 16 16 64 2 16 16 16

47 sotolong curry 2167 1103 - 32 128 64 32 64 32 32 8

48 syringole smoky, spices 2243 1344 - 8 32 8 16 8 128 8 1

49 (E)-isoeugenole floral, cloves 2326 1462 - 128 64 64 512 256 32 32 32

50 Z-6-dodecene-γ-lactoneg floral, stale 2379 1653 - - - 1 8 - - 1 -

51 unknown woody 2421 - 4 1 2 2 1 2 8 1

52 unknown dirty floral 2476 - - 1 2 1 4 1 1 -

60

Table 3.2 (cont.) Odor-active compounds identified by AEDA of nine rums

RIa Flavor Dilution Factorb,c

No. Compound Odor Descriptiond WAX RTX 5 BW BG AE RA7 AE12 DR12 ED12 RZ DX

53 ethyl vanilline vanilla 2492 1430 4 8 1 4 8 2048 16 4 2048

54 vanilline vanilla 2534 1396 128 128 128 1024 2048 4096 512 4096 256

55 unknown fresh/floral 2578 4 - 1 8 1 - 2 16 16

56 ethyl vanillatee vanilla 2606 1560 - 16 4 64 64 32 16 8 8

57 unknown band-aid 2850 - - - 2 4 4 4 - 2

58 syringaldehydee vanilla 2916 1656 2 16 64 32 16 64 8 32 16

59 unknown campfire 2951 1832 4 4 16 8 64 256 2 32 8 a Retention indexes determined from GCO data. b “BW” is Bacardi White, “BG” is Bacardi Gold, “AE” is Appleton Estate V/X, “RA7” is Ron Abuelo 7 year, “AE12” is Appleton Estate 12 year, “DR12” is Diplomatico Reserva Exclusica, “ED12” is El Dorado 12 year, “RZ” is Ron Zacapa Centurio, “DX” is Dictador XO Insolent. c FD factors were determined on a RTX-wax column and were determined from average log2 FD factors [n=2] after rounding up to the nearest whole number. d Odor properties determined by GCO. e Compound positively identified based on retention index on both RTX-wax and RTX-5 columns, odor property, mass spectral data, and reference standard compound. f Not detected. g Compound tentatively identified based on retention index on both RTX-wax and RTX-5 columns, odor property, and reference standard compound. h Compound tentatively identified based on retention index on RTX-wax column, odor property, and reference standard compound.

61

Figure 3.1 EI-Mass spectrum of unknown 59

62

3.6 References

Abe, E., Nakatani, Y., Yamanishi, T., & Muraki, S. (1978a). Studies on the “Sugary Flavor” of Raw

Cane Sugar. I. Proceedings of the Japan Academy. Ser. B: Physical and Biological Sciences, 54(B), 542–

547. http://doi.org/10.2183/pjab.54.542

Abe, E., Nakatani, Y., Yamanishi, T., & Muraki, S. (1978b). Studies on the “Sugary Flavor” of Raw

Cane Sugar. III. Proceedings of the Japan Academy. Ser. B: Physical and Biological Sciences, 54(B), 457–

462. http://doi.org/10.2183/pjab.54.542

Baldwin, S., Black, R. A., Andreasen, A. A., & Adams, S. L. (1967). Aromatic Congener Formation

in Maturation of Alcoholic Distillates. Journal of Agricultural and Food Chemistry, 15(3), 381–385.

http://doi.org/10.1021/jf60151a012



Puerto Rico, 62(4), 330–343.

Bober, A., & Haddaway, L. W. (1963). Gas Chromatographic Identification of Alcoholic Beverages.





Câmara, J. S., Alves, M. A., & Marques, J. C. (2006). Changes in volatile composition of Madeira

wines during their oxidative ageing. Analytica Chimica Acta, 563(1–2 SPEC. ISS.), 188–197.

http://doi.org/10.1016/j.aca.2005.10.031

Chatonnet, P., & Dubourdieu, D. (1998). Identification of Substances Responsible for the

“Sawdust” Aroma in Oak Wood. Journal of the Science of Food and Agriculture, 76, 179–188.

http://doi.org/10.1002/(SICI)1097-0010(199802)76:2<179::AID-JSFA924>3.0.CO;2-6

Conner, J. M., Paterson, A., & Piggott, J. R. (1993). Changes in Wood Extractives from Oak Cask

Staves through Maturation of Scotch Malt Whisky. Journal of the Science of Food and Agriculture, 62,

63

169–174. http://doi.org/10.1002/jsfa.2740620210

Cutzach, I., Chatonnet, P., Henry, R., & Dubourdieu, D. (1997). Identification of Volatile

Compounds with a “Toasty” Aroma in Heated Oak Used in Barrelmaking. Journal of Agricultural

and Food Chemistry, 45, 2217–2224. http://doi.org/10.1021/jf960947d

de Simón, B. F., Esteruelas, E., Muñoz, Á. M., Cadahía, E., & Sanz, M. (2009). Volatile Compounds

in Acacia, Chestnut, Cherry, Ash, and Oak Woods, with a View to Their Use in Cooperage.

Journal of Agricultural and Food Chemistry, 57, 3217–27. http://doi.org/10.1021/jf803463h



488. http://doi.org/10.1021/jf0511190

Fahrasmane, A., Parfait, A., Jouret, C., & Galzy, P. (1985). Production of Higher Alcohols and Short

Chain Fatty-Acids By Different Yeasts Used in Rum Fermentations. Journal of Food Science, 50,

1427–1436. http://doi.org/10.1111/j.1365-2621.1985.tb10492.x







Genthner, E. R. (2014). Identification of Rotundone as an Important Contributor to the Flavor of Oak Aged

Spirits. University of Illinois at Urbana Champaign. Retrieved from

https://www.ideals.illinois.edu/handle/2142/16898

Guth, H. (1997). Identification of Character Impact Odorants of Different White Wine Varieties.

Journal of Agricultural and Food Chemistry, 45, 3022–3026.

Kobayashi, A. (1989). Sotolon: Identification, Formation, and Effect on Flavor. In Flavor Chemistry:

Trends and Developments (ACS 388) (Vol. 388, pp. 49–59). American Chemical Society.

http://doi.org/10.1021/bk-1989-0388.ch005

64

Leading rum brands worldwide in 2015, based on sales volume (in million 9 liter cases). (2017).

Retrieved March 16, 2017, from https://www.statista.com/statistics/259739/leading-rum-

brands-worldwide-based-on-sales-volume/




Limited Ingredients. (2016). Alcohol and Tobacco Tax and Trade Bureau. Retrieved from

https://www.ttb.gov/ssd/limited_ingredients.shtml




Maga, J. A. (1989). The Contribution of Wood to the Flavor of Alcoholic Beverages. Food Reviews

International, 5(1), 39–99. http://doi.org/10.1080/87559128909540844

Matheis, K., Granvogl, M., & Schieberle, P. (2016). Quantitation and Enantiomeric Ratios of Aroma

Compounds Formed by an Ehrlich Degradation of L-Isoleucine in Fermented Foods. Journal of


Monsalve, J. O., Lopez, C., & Zapata, J. (2016). Characterization of aroma compounds in

Colombian rums by HS-SPME-GC-MS-O. Revista Colombiana de Química, 45(2), 48–54.

http://doi.org/http://dx.doi.org/10.15446/rev.colomb.quim.v45n2.60406.

Mosedale, J. R. (1995). Effects of oak wood on the maturation of alcoholic beverages with particular

reference to whisky. Forestry, 68(3), 203–230. http://doi.org/10.1093/forestry/68.3.203

Mosedale, J. R., & Puech, J.-L. (1998). Wood maturation of distilled beverages. Trends in Food Science

& Technology, 9, 95–101. http://doi.org/10.1016/S0924-2244(98)00024-7




65

Nordström, K. (1963). Formation of Esters from Acids by Brewer’s Yeast I. Kinetic Theory and

Basic Experiments. Journal of the Institute of Brewing, 69, 310–322.

Nykänen, B. L., Nykänen, I., & Suomalainen, H. (1977). Distribution of Esters Produced During

Sugar Fermentation Between the Yeast Cell and the Medium. Journal of the Institute of Brewing, 83,

32–34.

Nykänen, L., & Nykänen, I. (1977). Production of Esters by Different Yeast Strains in Sugar

Fermentations. Journal of the Institute of Brewing, 83, 30–31.

Nykanen, L., Puputti, E., & Suomalainen, K. (1968). Composition of the Aroma in Some Brands of

Whisky and Rum Analysed by Customary Methods and by Gas Chromatography. Kemian

Teollisuus, 25(5), 399–404.

Piggott, J. R., & Conner, J. M. (2003). Whiskies. In A. G. . Lea & J. R. Piggott (Eds.), Fermented

Beverage Production (2nd ed., pp. 239–262). New York: Kluwer Academic/Plenum Publishers.




Pino, J. A., Marbot, R., Perez, J., & Nunez de Villavicencio, M. (1999). Application of Multivariate

Statistical Methods to Differentiation of Rum Aged in Two Types of Oak Barrels. Journal of

Food Quality, 22, 183–191. Retrieved from %5C%5CRobsrv-05%5Creference

manager%5CArticles%5C8919.pdf






954, 51–57.

Poisson, L., & Schieberle, P. (2008). Characterization of the Most Odor-Active Compounds in an

American Bourbon Whisky by Application of the Aroma Extract Dilution Analysis. Journal of

66

Agricultural and Food Chemistry, 56, 5813–5819.

Pons, A., Lavigne, V., Landais, Y., Darriet, P., & Dubourdieu, D. (2010). Identification of a Sotolon

Pathway in Dry White Wines. Journal of Agricultural and Food Chemistry, 58, 7273–7279.

http://doi.org/10.1021/jf100150q

Sefton, M. a, Skouroumounis, G. K., Elsey, G. M., & Taylor, D. K. (2011). Occurrence, Sensory

Impact, Formation, and Fate of Damascenone in Grapes, Wines, and Other Foods and

Beverages. Journal of Agricultural and Food Chemistry, 59, 9717–46.

http://doi.org/10.1021/jf201450q

Silvia Ferreira, A. C., Barbe, J., & Bertrand, A. (2003). 3-Hydroxy-4,5-dimethyl-2(5H)-furanone: A

key Odorant of the Typical Aroma of Oxidative Aged Port Wine. Journal of Agricultural and Food

Chemistry, 51, 4356–4363. http://doi.org/10.1021/jf0342932

Stevens, R., & Martin, G. (1965). Gas Chromatographic Identification of Ethyl Esters of Fatty Acids

in Domestic and Imported Rums. Journal of the Association of Official Agricultural Chemists, 48(4),

802–805.



Thuy, P. T., Elisabeth, G., Pascal, S., & Claudine, C. (1995). Optimal Conditions for the Formation

of Sotolon from α-Ketobutyric Acid in the French “Vin Jaune.” Journal of Agricultural and Food

Chemistry, 43(10), 2616–2619. http://doi.org/10.1021/jf00058a012



Tokitomo, Y., Kobayashi, A., Yamanishi, T., & Muraki, S. (1980). Studies on the “Sugary Flavor” of

Raw Cane Sugar. II. Proceedings of the Japan Academy. Ser. B: Physical and Biological Sciences, 56, 452–

456.

van Den Dool, H., & Kratz, P. D. (1963). A Generalization of the Retention Index System Including

Linear Temperature Programmed Gas—Liquid Partition Chromatography. Journal of

Chromatography, 11, 463–471.

67



68

Chapter 4: Quantitation of Odor-Active Compounds in a Selection of

Premium Rums and Use of Chemometrics to Correlate Analytical

and Sensory Analysis

4.1 Abstract

Thirty-four of 50 odor-active compounds identified in the previous study in mixing and premium

rums were quantitated. Results revealed the same compounds were present in all rums analyzed,

except the Bacardi White rum and the compound ethyl vanillin; however, they varied in their

concentration accounting for flavor differences among rums. Sixteen compounds were determined

to be the most important components of overall rum flavor since they were detected in all nine rum

samples with odor activity values > 1. These consisted of 2-methyl propanal, acetal, 3-methyl

butanal, 2-methyl butanal, ethyl 2-methylpropanoate, ethyl butanoate, ethyl 2-methylbutanoate, ethyl

3-methylbutanoate, 3-methyl-1-butanol, 2-methyl-1-butanol, ethyl hexanoate, β-damascenone,

guaiacol, cis-oak lactone, and vanillin. Premium rums tended to have higher concentrations and

subsequently higher OAVs than mixing rums for most compounds, except for Dictador XO

Insolent 21 year rum, in which concentrations for some compounds were lowest among all rums.

Chemometrics was employed to gain a better understanding of the drivers of aroma variations in the

rums analyzed. Vanillin and ethyl vanillate, were found to be driving factors of sweet-like aroma

attributes including brown sugar, caramel, maple and vanilla aromas. Meanwhile, roasted aroma was

driven by a decrease of several compounds rather than by an increase or decrease in any single

odorant.

4.2 Introduction

Rum is a complex and diverse spirit, produced from sugarcane by-products including sugarcane

juice, syrup or molasses. Rum’s minimum standard of identity allows for a variety of manufacturing

practices to be used in its production. Key differences in production styles include variations in

staring material, yeast and bacteria used for fermentation, length of fermentation, distillation

69

apparatus, aging system, barrel type and maturation length. The resulting rums span from light one-

year aged white rums, to heavy Jamaican style 50 year old rum. The breadth of rum types available

make it difficult to identify the key constituents of overall rum flavor. Previous studies on rum flavor

have only focused on the evaluation of one or two rum samples, making it difficult to define the

entire category (Burnside, 2012; de Souza, Vasquez, Del Mastro, Acree, & Lavin, 2006; Franitza,

Granvogl, & Schieberle, 2016a, 2016b; Pino, Tolle, Gök, & Winterhalter, 2012). Additionally, only

three of these studies have quantitated the odor-active compounds identified in rum (Franitza et al.,

2016a, 2016b; Pino et al., 2012) and only the studies by Franitza have used stable isotope dilution

analysis for compound quantitation. In our previous study (Chapter 3), the key aroma compounds

were identified for a variety of premium rums and a set of mixing rums on the basis of flavor

dilution factors determined by application of aroma extract dilution analysis. A total of 59 odor-

active regions, encompassing 50 identifiable compounds and 14 unknown regions.

Chemometrics is the use of statistics to analyze chemical data (Beebe, Pell, & Seasholtz, 1998).

Chemometrics can be used in one of two ways, either by creating a calibration for use in predicting

and/or identifying future samples or to help identify patterns in complex datasets (Beebe et al.,

1998; Roberts & Cozzolino, 2016). A number of different methods exist for both types of analyses.

Model creation methods include artificial neural networks, multiple linear regression, principle

components regressions and partial least squares discrimination analysis. Pattern recognition

analysis includes hierarchical cluster analysis and principal components analysis (PCA). Partial least

square discriminant analysis and PCA are by far the most often methods used in the correlation of

sensory and analytical data (Seisonen, Vene, & Koppel, 2016).

In the present study, principal components analysis was selected, as the goal of this research was to

evaluate the relationship between sensory and analytical data, and not to create a model system. The

purpose of PCA is to reduce the number of variables to the fewest number of factors possible

(Beebe et al., 1998; Brereton, 2007, 2009). These new factors are then plotted instead of the original

variable measurements allowing for easier visualization and interpretation of the data.

Chemometrics has been used extensively to evaluate food systems and to better understand the

complexity of food matrices as illustrated by Roberts and Cozzolino (2016) in their recent review

article. Focusing specifically on flavor chemistry, a significant number of studies have used

chemometrics to correlate sensory and analytical data (Seisonen et al., 2016). Alcoholic beverages

70

that have been investigated using chemometrics include beer (da Silva et al., 2012; Dong et al., 2014),

whiskey (Wiśniewska, Śliwińska, Dymerski, & Wardencki, 2017), and rice wine (Jung, Lee, Lim, Kim,

& Park, 2014; Mimura, Isogai, Iwashita, Bamba, & Fukusaki, 2014), with the majority of studies

focused on wine (Andreu-Sevilla, Mena, Martí, García Viguera, & Carbonell-Barrachina, 2013;

Aznar, López, Cacho, & Ferreira, 2003; González Álvarez, González-Barreiro, Cancho-Grande, &

Simal-Gándara, 2011; Green, Parr, Breitmeyer, Valentin, & Sherlock, 2011; Liu et al., 2015; Niu et

al., 2011; Robinson et al., 2011; Vilanova, Genisheva, Masa, & Oliveira, 2010; Xiao et al., 2014).

Chemometrics has also been used to profile rum (Sampaio, Reche, & Franco, 2008) and cachaça, a

Brazilian sugarcane spirit similar to rum (Granato, de Oliveira, Caruso, Nagato, & Alaburda, 2014).

Most studies focused on the differentiation of the two beverages (Cardoso et al., 2004; P. P. de

Souza et al., 2007; Todeschini et al., 2007). These studies used mineral concentration, electrospray

ionization-mass spectrometry, and concentrations of selected compounds consisting mainly of

higher alcohol and phenolic compounds as variables to differentiate the rum and cachaça samples.

However, to date chemometrics has not been implemented to correlate the sensory attributes and

volatile aroma components of rum or cachaça.

The goal of this research was two-fold. The first aim was to first quantitate the odor-active

compounds identified in our previous study (Chapter 3). The second aim was to use chemometrics

to correlate the analytical data (concentrations, odor activity values, and flavor dilutions factors) with

sensory aroma intensity ratings for the same rum samples determined in a previous study (Chapter

6) in order to gain a better understanding of how sensory attributes are affected by changes in

concentration of odor-active compounds.


Materials

The nine selected rums were purchased at a local liquor store (Champaign, IL) (Table 3.1). All nine

rums had reported ethanol concentration of 40% alcohol by volume (ABV). Mention of the brand

name of these rums does not imply any research contact or sponsorship, and is not for

advertisement or endorsement purposes.

71

Chemicals

Dichloromethane, ethyl acetate, hydrochloric acid, calcium chloride, potassium hydroxide, sodium

chloride and sodium sulfate were purchased from Fisher Scientific Co. (Fair Lawn, NJ); 3,4-

dihydroxy-5-methoxybenzaldehyde, d6-dimethylfulate and d6-ethanol were purchased from Aldrich

(Milwaukee, WI) and used for volatile extraction and synthesis of isotopic standards. De-odorized

water was prepared by boiling deionized-distilled water to two-thirds of its original volume.

Deodorized water and 190 proof ethanol (Decon Labs, Inc. USP grade, King of Prussia, PA) were

used to create the 40% ABV mimic matrix.

Standard Compounds

The following chemicals were used as authentic standards to determine response factors for

quantitation with isotopes and relevant odor thresholds: 2-methylpropanal, ethyl propanoate, ethyl-

3-methylbutyrate, isoamyl acetate, ethyl hexanoate, ethyl octanoate, p-cresol, and eugenol, were

obtained from Aldrich (Milwaukee, WI); acetaldehyde, 3-methylbutanal, ethyl 2-methylpropanoate,

ethyl butyrate, ethyl-2-methylbutyrate, ethyl pentanoate, 2-methyl-1-butanol, 3-methyl-1-butanol,

mixture of cis-&-trans whiskey lactone, 2-phenethyl alcohol, guaiacol, syringol, isoeugenol, and ethyl

vanillin were obtained from Sigma-Aldrich (St. Louis, MO); acetic acid, 4-methylguaiacol, 4-

ethylguaiacol, m-cresol, sotolon, vanillin, syringaldehyde were obtained from SAFC (St. Louis, MO);

ethyl vanillate was obtained from Alfa Aeasar (Lancaster, UK); 2-methylbutanal was obtained from

Bedoukian (Danbury CT); acetal (1,1-diethoxyethane) was obtained from Acros Organics (NJ);

isobutanol was obtained from Fisher (Fair Lawn, NJ); and β-damascenone was obtained from

Firmenich (Switzerland).

Isotopes for Quantitation

The following compounds were purchased to be used as isotopically labeled standards quantitation:

d7-ethyl butanoate, d11-ethyl hexanoate, d3-guaiacol, and d3-p-cresol were purchased from C/D/N

Isotopes Inc., (Pointe- Claire, Quebec, Canada), and d8-m-cresol was purchased from Sigma-Aldrich

(St. Louis, MO).

Isotopically labeled standards not available for purchased were synthesized according to the

published procedure in parentheses: d2-2-methyl propanal, d2-3-methyl butanal, d2-2-methylbutanal

72

(Lapsongphon, Yongsawatdigul, & Cadwallader, 2015; Steinhaus & Schieberle, 2005) d2-isoamyl

acetate, d2-cis- & trans-whiskey lactone, d2-4-methylguaiacol, d3-vanillin, d5-ethyl vanillate (Lahne,

2010), d2-isobutanol (Lahne, 2010; Steinhaus & Schieberle, 2005), d2-3-methyl-1-butanol (Steinhaus &

Schieberle, 2005), d2-2-methyl-1-butanol (Kelley, 2014; Steinhaus & Schieberle, 2005), d4-ethyl

octanoate (Genthner, 2014), d4-β-damascenone (Kotseridis, Baumes, & Skouroumounis, 1998;

Lahne, 2010), 13C2-2-phenethyl alcohol (Lahne, 2010; Schuh & Schieberle, 2006), d5-4-ethylguaiacol

(Lahne, 2010; Rayne & Eggers, 2007), d3-eugenol (Kulkarni, Kadam, Mane, Desai, & Wadgoankar,

1999; Lahne, 2010; Schneider & Rolando, 1992), d3-syringol (Lahne, 2010; Schneider & Rolando,

1992), d3-(E)-isoeugenol (Lorjaroenphon, 2012), d10-acetal and d3-syringaldehyde (below).

Structures for all labeled isotopes mentioned above are shown in Figure 4.1.

Synthesis of d10-Acetal

d10-acetal was synthesized following the procedure by Adkins and Nissen (1941). d6-Ethanol

(1.26mL) and anhydrous calcium chloride (0.2g) were added to a 50mL screw top test tube and

cooled in an ice bath. Next, freshly distilled acetaldehyde (0.46mL) was slowly pipetted down the

side of the tube, to form a layer on top of the alcoholic calcium chloride. The test tube was capped

and shaken vigorously for 2 minutes. The mixture was then allowed to stand for one day at room

temperature with intermittent shaking. The reaction was checked by GC-MS to determine that it

had gone to completion. Once the reaction was complete, the top layer (acetal) was pipetted off

into a separatory funnel, and the acetal was washed 3x with 2mL of water.

Synthesis of d5-Ethyl Esters

d5-Ethyl propanoate was synthesized as follows. Propionic acid (0.74 g; 10 mmol), d6-ethanol (0.158

g, 3 mmol; Aldrich, 99.5% atom %D) and 2 drops of concentrated H2SO4 were added to a 22-mL

glass vial. The vial was sealed with a PTFE-lined silicon cap and then heated at 100C for 2 h.

After cooling the vial to room temperature the reaction mixture was diluted 10 mL of pentane and

extracted with a saturated aqueous Na2CO3 solution (2 x 5 mL). The pentane layer was washed with

saturated aqueous NaCl (2 x 5 mL) and dried over anhydrous Na2SO4. The target compound (0.21

g) was recovered after removal of the pentane using a gentle stream of N2 gas:

73

d5-Ethyl 2-methylbutanoate, d5-ethyl 2-methylpropanoate, d5-ethyl 3-methylbutanoate were

synthesized following the same procedure.

Synthesis of d3-Syringaldehyde

d3-Syringaldehyde was synthesized following the procedure by Lahne (2010). First 0.501 g (3 mmol)

of 3,4‐dihydroxy-5-methoxybenzaldehyde was dissolved in aqueous 40% KOH (5mL) under a

nitrogen purge in a screw-capped test tube (PTFE-top) equipped with a stir bar. Then, over the

course of 30 minutes (5-6 drops every 5 minutes), 0.35 mL (0.42 g, 3.2 mmol) of d6‐dimethylsulfate

was added to the reaction tube after which the reaction mixture became yellow and cloudy. The vial

was then capped and stirred for 2 hours. The reaction was checked for completion by removing 5-6

drops of the mixture, adding it to a vial containing 1 mL aqueous 1N HCl and 0.5 mL ethyl acetate,

and analyzing the ethyl acetate layer by GC-MS. The reaction was continued, adding 0.08 mL (0.096

g, 0.73 mmol) d6-dimethylsulfate and letting the reaction stir overnight until nearly all starting

material had been consumed. The reactions was stopped by acidifying the mixture to ~pH 1 and

extracted with ethyl acetate (1 x 10mL, 4 x 5 mL). The ethyl acetate layer was washed with saturated

NaCl and then dried over anhydrous Na2SO4. The solution was concentrated to ~10 mL using a

vigreux column and the remaining solvent was then removed under a stream of nitrogen. The final

product was weighed for a final yield of 0.5734 g.

Methods

Stable Isotope Dilution Analysis (SIDA)

Isotopically labeled (2H or 13C) standard compounds were used for the quantitation of compounds

determined to be potent odorants by AEDA. For analysis by direct solvent extraction, the rum

samples (10 mL) were spiked with a known concentration of the isotope for each compound of

interest. Next, deodorized water (30 mL) was added to dilute the alcohol concentration to

~10%ABV and the samples were then shaken for 5 minutes. Dichloromethane (2 mL) was added to

each sample to extract the isotopes and target compounds. The samples were shaken for an

additional 5 minutes and then centrifuged at 3500 RPM to separate the aqueous and organic layers.

The bottom layer (dichloromethane) was transferred to a 15 mL conical test tube. The sample was

extracted two more times and the extracts were pooled. The extract was then dried over sodium

74

sulfate and the sample concentrated to 0.5 mL under a gentle stream of nitrogen. The samples were

stored at -20 °C until analyzed.

GC-MS Analysis

Solutions for analysis by GC-MS were prepared for both mass spectra and labeled:unlabeled mass

ratio calibration. Each solution was analyzed by GC‐MS in cold-splitless mode (-50°C for 0.10 min,

then increased at 12°C/sec to 260°C, with 1.10 min valve-delay), using the same stationary columns

(Stabiwax column and RTX-5 column) as for the analysis of the samples.

The GC-MS system used for quantitation consisted of a 6890 GC/5973N mass selective detector

(Agilent Technologies Inc.). Two μL of spiked extract was injected into a CIS-4 inlet (Gerstel,

Germany) in the cold splitless mode (-50°C for 0.10 min, then increased at 12°C/sec to a final

temperature of 260°C). Separations were performed using a Stabilwax column (30 m length x 0.25

mm i.d. x 0.25 μm film thickness; Restek; Bellefonte, PA). Helium was used as the carrier gas at 0.7

mL/minute. Oven temperature was programmed as follows: initial temperature, 35 °C (5 min hold),

ramp rate 4 °C/min, final temperature, 225 °C (30 min hold). For each compound the key

identifying ions were scanned for during the run using selection ion monitoring (SIM) mode. Table

4.1 contains the list of compounds quantitated along with their ions analyzed and the response

factor from the standard curve.

For each compound a response factor was determined by plotting the ratio of the mass of the

standard of interest / mass of corresponding isotope vs the ratio of the area count of the standard/

area count of the isotope. The response factor was then calculated from 1 over the slope:

𝑅𝑓 = 1

𝑠𝑙𝑜𝑝𝑒

For each compound, the area of the selected mass ion on the chromatogram was integrated using

Enhanced Data Analysis Software (Agilent Technologies, USA). The mass ratio of the

labeled:unlabeled compound is plotted against the chromatogram area of the selected mass ion,

using the slope of the line to calculate the response factor for each compound. Standard curve data

from all compounds quantitated can be found in Appendix A. The response factor used to calculate

the actual concentration of the compound in the sample using the following equation:

75

𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛𝑢𝑛𝑙𝑎𝑏𝑒𝑙𝑒𝑑 = 𝑎𝑟𝑒𝑎𝑢𝑛𝑙𝑎𝑏𝑒𝑙𝑒𝑑

𝑎𝑟𝑒𝑎𝑙𝑎𝑏𝑒𝑙𝑒𝑑∗ 𝑅𝑓 ∗ 𝑚𝑎𝑠𝑠𝑙𝑎𝑏𝑒𝑙𝑒𝑑 𝑠𝑎𝑚𝑝𝑙𝑒 𝑣𝑜𝑙𝑢𝑚𝑒⁄

where the area of the target compounds is divided by the area of the isotope, then multiplied by the

response factor, multiplied by the mass of the isotope in the samples (determined by multiplying the

volume of isotope spiked by the concentration of the solution) and finally divided by the volume of

sample initially spiked with the isotope.

Headspace Solid-Phase Microextraction (HS-SPME)

Headspace solid-phase microextraction (HS-SPME) was used to quantitate the highly volatile aroma

compounds. Rum (0.5 mL) was pipeted into a 20 mL glass headspace vial containing 0.5 g of

sodium chloride and 1.5 mL of deodorized water. The samples were then spiked with the isotope

solutions. Samples were analyzed by GC-MS using an auto-sampler. Samples were equilibrated at

60°C for 10 minutes and then the headspace volatiles were extracted by placing a triple phase SPME

fiber (divinylbenzene/carboxen/polydimethylsiloxane) (Supelco; Bellefonte, PA) to extract the

volatiles for 5 minutes at 60°C. The SPME fiber was then removed and the volatiles desorbed into a

Gerstal CS4 injection port (splitless injection at 260°C with 4 minute valve-delay).

Quantitation of Strecker Aldehydes

Strecker aldehydes (methylpropanal, 3-methylbutanal, and 2-methylbutanal) were quantitated by

SIDA using HS-SPME-GC-MS. The conditions for SPME extraction and desorption were the same

as above. The GC-MS parameters were as follows: separations were performed using an RXi-5ms

column (30 m length x 0.25 mm i.d. x 0.25 μm film thickness; Restek; Bellefonte, PA). Helium was

used as the carrier gas at 1.0 mL/minute. Oven temperature was programmed as follows: initial

temperature, 30° C (5 min hold), ramp rate 6° C/min, final temperature, 225° C (30 min hold).

Other GC-MS conditions were the same as earlier described.

Quantitation of Acetal

Additionally, acetal was quantitated using direct injection methodology. The rum samples (1 mL)

were spiked with a known concentration of the isotope immediately before injection and then

shaken for one minute. Then, a 1 µL sample was injected into the GC-MS using hot, splitless

injection (260°C with 1.10 minute valve-delay) and separations performed using a RTX-5 column.

76

(30 m length x 0.25 mm i.d. x 0.25 μm film thickness; Restek; Bellefonte, PA). Helium was used as

the carrier gas at 0.5 mL/minute. Oven temperature was programmed as follows: initial temperature,

35° C (5 min hold), ramp rate 6° C/min, final temperature, 225° C (30 min hold). Other GC-MS

conditions were the same as earlier described.

Quantitation of Ethyl Vanillin

Ethyl vanillin was quantitated using internal standard methodology. The rum samples were extracted

following the same procedure used for SIDA, spiking the samples with a known concentration of d3-

vanillin. A standard curve comparing mass ratios and area ratios for selected ions of ethyl vanillin to

d3-vanillin were constructed and can be found in Appendix A.

Quantitation of Acetaldehyde and Acetic Acid

Acetaldehyde and acetic acid were quantitated using external standard calibration. The GC-FID

system used for quantitation consisted of a 6890N GC (Agilent Technologies Inc.) equipped with a

flame ionization detector (FID). Two μL of straight rum was injected using a 7683 autosampler and

injector (Agilent Technologies Inc.) into a split/splitless inlet in the hot split mode (260° C, 5:1 split

ratio). Separations were performed using a Stabiwax-DA column (30 m length x 0.32 mm i.d. x 0.5

μm film thickness; Restek; Bellefonte, PA). Helium was used as the carrier gas at 2.0 mL/minute.

FID temperature was 250°C. Oven temperature was programmed as follows: initial temperature, 35

°C (5 min hold), ramp rate 10 °C/min, final temperature, 225 °C (20 min hold).

Calibration solutions were prepared by spiking known amounts of acetaldehyde or acetic acid into a

40% ABV mimic matrix. Standard curves for acetaldehyde and acetic acid can be found in Appendix

A.

Determination of Odor Thresholds

All procedures and recruitment material was approved by the Institutional Review Board (IRB) at

the University of Illinois Urbana-Champaign (IRB Protocol Number: 17508), Appendix B.

Odor thresholds of ethyl vanillate, isoeugenol, and syringaldehyde were determined in 40% ABV

ethanolic matrix. Odor purity of the samples was checked prior to threshold determination by

77

running the most concentrated dilution by GCO. Samples were determined odor pure if no other

odorants were detected at that concentration.

A series of seven samples was prepared for sensory evaluation for each compound. A stock solution

of each compound was prepared by dissolving the compounds of interest in 100 mL a 40% ABV

ethanolic matrix. The solution was pipetted (20 mL) into a 125 mL Teflon bottle. A 1:3 dilution was

prepared by adding 20 mL of 40% ABV ethanolic matrix to a Teflon bottle and then adding 10 mL

of the stock solution. The remaining five solutions in the series were prepared in the same manner,

taking 10 ml of the previous solution for the dilution. The final volume in each bottle was 20 mL.

Odor thresholds were determined by sensory evaluation through the use of the 3-alternative forced

choice (3-AFC) method, according to the American Society for Testing and Materials (ASTM)

method for determining odor thresholds (ASTM E679-04, 2011) using an ascending concentration

series. A minimum of seven panelists were used for each threshold determination. For each

concentration level, panelists were presented with one bottle containing the compound of interest

and two blanks (20 ml 40% ABV ethanolic matrix in a Teflon bottle). Panelists started with the least

concentrated sample and evaluated samples in ascending order. Samples were covered with

aluminum foil and labeled with a random three-digit code prior to analysis. Panelists were asked to

sniff the bottles in the order indicated on the sheet and then select which of the samples was

stronger than the other two. Panelists evaluated the series in duplicate with a five-minute break

between replications. The concentrations used for each dilution series and panelists results can be

found in Appendix C.

Statistical Analysis of Quantitation Data

Statistical analysis of the data was performed using Statistical Analysis System (SAS)® (Version 9.4,

SAS Institute Inc., Cary, NC, USA). Analysis of variance (ANOVA) was conducted on each

compounds quantitated to determine the presence of overall significant differences (p<0.05) using

the PROC GLM function for variations aroma concentration between the nine rums. The calculated

probabilities were compared to significance levels α=0.05, 0.01 and 0.001. Fisher’s least significant

difference (LSD) test was conducted on all attributes determined as significant by ANOVA. Cluster

analysis was also conducted using SAS software using the PROC CLUSTER function on the mean

quantitation and OAV data separately.

78

Chemometric Analysis

Principle component analysis (PCA) was employed to compare the analytical and sensory results.

Sensory scores for the aroma attributes evaluated in the nine rum samples obtained from previous

descriptive analysis panel (Chapter 6) were used. The sensory scores were compared to multiple

analytical results: concentrations, odor activity values and FD factors (Chapter 3). Odor activity

values and FD factors were pre-process prior to evaluation. OAV were pre-processed by removing

all compounds that had OAVs less than 1 for all rum samples. For FD factors, only odor-active

regions that contained at least one FD factor of 8 or higher were used for statistical analysis.

Statistical analysis was performed using SAS. Principle component analysis (PCA) was conducted

using the PROC CORR function followed by the PROC FACTOR function. Microsoft® Excel®

2016 (Version 16: Microsoft Corporation, Redmond, WA) was used to create a visual representation

of the data to allow further examination of the relationship of the rums to individual attributes that

characterized the samples. Pearson correlations were calculated using the same SAS software, with

significance determined at α=0.1, 0.05, and 0.01.


Quantitation of Potent Odorants in Mixed and Premium Rum

Stable isotope dilution analysis (SIDA) was used for accurate quantitation of selected odor-active

compounds in rums. SIDA is a highly accurate method of quantitation for volatile compounds as

isotopically labeled compounds are used as the internal standards. The isotopes differ only from the

target compounds by the replacement of carbon atoms (12C) with heavy carbon atoms (13C) or by

replacement of hydrogen atoms (1H) with deuterium atoms (2H). As a result, the physical and

chemical properties of the isotopes are the same as the corresponding standard. Therefore, the

isotope and standard will interact with the matrix and extraction solvents in the same way. Losses of

the two compounds will be the same, and the ratio established when the internal standard is added

to the sample will be the same as when the extract is analyzed. All isotopes used in this study (Figure

4.1) were labeled with deuterium except for 2-phenethyl alcohol which was labeled with heavy

carbon (13C).

79

A total of 34 odor-active compounds positively identified in the previous study (Chapter 3) were

selected for quantitation by SIDA or external standard calibration (Table 4.2). All compounds

quantitated were found in all nine rum samples regardless if they were detected by AEDA with the

exception of 4-ethylguaiacol and eugenol, which were not detected in BW. This is to be expected as

they are compounds generated during wood aging (Cutzach, Chatonnet, Henry, & Dubourdieu,

1997; Maga, 1989). Additionally, ethyl vanillin was only detected in DR12 and DX. ANOVA was

performed for each compound quantitated, revealing significant differences in concentration across

the nine rums for all compounds. Calculated F-values can be found in (Table 4.3), and mean

separation by Fisher’s Least Significant Difference test was performed on all compounds with

concentrations found to be statistically different between rums.

The most abundant compounds in all nine rums were 3-methyl-1-butanol, isobutanol, acetaldehyde,

acetic acid, 2-methyl-1-butanol and acetal. This is in agreement with the previous studies by Franitza

(2016a, 2016b), who also found these compounds to be the most abundant in rum, with the

exception of acetaldehyde which was not quantitated in that study. Interestingly, while these

compounds varied in concentration among the rums, the relative order of abundance was relatively

similar. In BW, BG, AE and RZ the compounds descended in concentration in the order as listed

above. RA7, AE12 and ED12 had the same order of descending compound except acetic acid was

the most abundant compound in these three rums. The order of abundance for DR12 and DX were

similar to the concentration order of RA7, AE12, and ED12 except the order of 2-methyl-1-butanol,

acetaldehyde, and acetal varied. Aside from these compounds, the order of other compounds varied

among rums. Syringaldehyde, 2-phenethyl alcohol, cis-whiskey lactone, vanillin and ethyl propanoate

appeared as the next 7 most abundant compounds in all nine rums. The least abundant compounds

in all samples tended to be 4-ethylguaiacol, 4-methylguaiacol, m-cresol, and p-cresol. Other

compounds present in low abundance included β-damascenone, eugenol, ethyl 2-methylbutanoate,

and ethyl pentanoate.

Ethyl vanillin was only detected in two samples, DR12 and DX. Ethyl vanillin is a synthetic flavor

compound commonly used as artificial vanilla and its presence in rum was not expected. However,

the Alcohol and Tobacco Tax and Trade Bureau (TTB) lists ethyl vanillin as one of four compound

on its limited ingredients list (Limited Ingredients, 2016). Compounds on this list can be added to

distilled beverages without having to claim them on the label or identifying the product as imitation.

As long as ethyl vanillin is present at concentrations less than 16 ppm, and the “total vanillin”

80

concentration (sum of the concentration of vanillin plus 2.5 times the ethyl vanillin concentration) is

less than 40 ppm the final rum can still be called a natural product. The “total vanillin”

concentrations of both DX and DR12 are well under this limit, 8.75 ppm and 5.24 ppm respectively.

Comparing compounds across rums, AE12 and ED12 has the highest concentrations for the

majority of odorants. Overall, the aged rums tended to have the highest concentration of all

compounds with the exception of DX. BW, BG and DX which consistently had the lowest

concentrations of all compounds quantitated. Interestingly, DX had the highest concentration of

ethyl butanoate and β-damascenone among all nine rums. It is difficult to say why the compound

concentration of DX is so low in comparison to the other aged rums. The low concentrations in DX

might be a result of maturation by the solera system or the extended aging period (21 years) of the

rum, as many of the volatiles may have been lost to evaporation over time. A more detailed

investigation into the production process of DX would be required to better explain the overall low

compound abundance in this rum.

BW and BG were expected to have the lowest concentration of many of the compounds This is due

to the fact that the two mixing rums spent the least amount of time in barrels and were distilled by

continuous column distillation which removes more fermentation-derived odor-active compounds

compared with other distillation methods such as pot distillation. It was hypothesized that the

mixing rums may have higher concentrations of the more volatile compounds such as acetaldehyde

or the Strecker aldehydes that may be lost due to evaporation over time in the aged rums but this

was not the case.

Cluster analysis was performed using the quantitations data (Figure 4.2). BG and BW were found to

be the most similar to each other as the distance between the two rums is smallest of any cluster

pairing and less than half of the distance to the next cluster. As was expected, the aged rums (RA7,

AE12, DR12, ED12, and RZ) were grouped together, with the exception of DX, which was

grouped with the younger aged rums (BW, BG, AE). Groupings within the aged rums revealed no

other significant correlations between quantitation data and the disclosed production processes

(Table 3.1). The grouping of DX with the younger rums in most likely related to the fact that overall

the concentration for most compounds was significantly lower than the other aged rums.

81

Calculation of OAVs

Calculation of odor activity values (OAVs) is necessary to better understand the impact of odorants

on the final rum flavor. While compound concentration reveals how much is present in the sample,

it gives no indication as to how important the compound is to the overall aroma. Some compounds

may be present in high concentrations but have a minimal impact on the overall aroma while other

compounds may be low threshold potent odorants, where they have low concentrations but high

impact factors. An odor detection threshold is the concentration a compound required for it to be

perceived in a matrix. Since compounds will react differently in different matrices, it is best to

determine thresholds in a matrix as similar to the final product as possible. Research has shown that

odor thresholds do change with changes in ethanol concentration. Therefore, thresholds determined

in a 40:60 (v/v) ethanol/water matrix were used when possible. OAVs are calculated by dividing the

concentration of the compound in the matrix by the concentration needed to detect the compound

in the matrix. Compounds with OAVs over 1 are assumed to impact the overall flavor perception of

the food or beverage.

The calculated OAVs for the nine rum samples are presented in Table 4.4. The number of

compounds with OAVs greater than 1 varied among rums. The number of compounds present at

concentrations above their thresholds was 17 for BW, 19 for BG, 23 for AE, 23 for RA7, 25 for

AE12, 23 for DR12, 22 for ED12, 24 for RX, and 19 for DX. The order of compound potency

varied among rums as well as OAVs for a single compound across all nine rum samples. For

example, the OAV of 2-methylpropanal in AE12 was 488 whereas it was only 6 in DX. Significant

difference is OAVs were also seen among rums for 3-methylbutanal (OAVs of 12 to 432), ethyl 2-

methylpropanoate (OAVs of 3 to 154), ethyl 2-methylbutanoate (OAVs of 18 to 369), ethyl 3-

methylbutanoate (OAVs of 9 to 149), β-damascenone (OAVs of 34 to 439), and vanillin (OAVs of

15 to 217).

Of the 34 compounds quantitated, 26 of the compounds had OAVs greater than 1 in at least one of

the rums. Fifteen of the compounds had OAVs greater than 1 in all nine samples, specifically: 2-

methylpropanal, acetal, 3-methylbutanal, 2-methylbutanal, ethyl 2-methylpropanoate, ethyl

butanoate, ethyl 2-methylbutanoate, ethyl 3-methylbutanoate, 3-methyl-1-butanol, 2-methyl-1-

butanol, ethyl hexanoate, β-damascenone, guaiacol, cis-whiskey lactone, and vanillin. When excluding

the mixing rums, ethyl octanoate and eugenol also have OAVs above 1 for all premium rums.

82

Additionally, these compounds contained the highest OAV values across all rum samples as all

compounds with OAVs of 10 or greater were among these 17 compounds. The most potent

odorants were β-damascenone for BG (OAV 90), RA7 (OAV 272), DR12 (OAV 143), and DX

(OAV 439), 2-methylpropanal for BW (OAV 52), AE (OAV 400) and AE12 (OAV 488), 3-

methylbutanal for ED12 (OAV 288) and ethyl 2-methylpropanoate (OAV 154).

4-Methylguaiacol, p-cresol, m-cresol, ethyl vanillate and syringaldehyde all had OAVs less than 0.01

in all rums. Syringol, trans-whiskey lactone and (E)-isoeugenol had OAVs less than 1 for all rums.

While these compounds may not have a direct impact on the aroma of rum, compounds with high

abundance such as syringaldehyde, ethyl vanillate, (E)-isoeugenol and trans-whiskey lactone, may

affect the partitioning of other compounds into the headspace and may be necessary for the creation

of an accurate model system.

The variation in odorants with OAVs greater than 1 is likely the cause of the differences perceived

in rum aroma within the category. While there seems to be a core set of compounds important to all

rums, the relative differences in concentration differentiate them from each other. The only

compounds that specifically defines a set of rums is ethyl vanillin in DR12 and DX. The inclusion of

ethyl vanillin in these samples is likely the result of the addition to the rums during the

manufacturing process as discussed previously. The most significant differences observed between

premium and mixing rums were the OAVs for cis-whiskey lactone and vanillin. Noticeable

differences also existed for ethyl octanoate and acetic acid.

Cluster analysis was also performed on the nine rums using the OAV results (Figure 4.3).

Interestingly, the rums were grouped differently than when using the quantitation data (Figure 4.2).

BW and BG are still grouped together as the most similar to each other, with the smallest distance

between clusters. AE and AE12 are also grouped together, although while they are grouped as more

similar to each other, they are not as closely related as the other seven rums are to themselves as

indicated by the distance between clusters. The aged rums are no longer clearly separated from the

other rums and DX is grouped much farther away from the mixing rums despite its overall low

concentration of volatiles.

83

Comparison to previous studies

Of the 34 compounds quantitated, 24 had previous been quantitated in rum after being identified as

odor-active constituents (Franitza et al., 2016a, 2016b; Pino et al., 2012). Ten compounds,

acetaldehyde, 2-methylpropanal, ethyl propanoate, 4-methyl guaiacol m-cresol, syringol, isoeugenol,

ethyl vanillin, and syringaldehyde were quantitated for the first time in rum. However, only

acetaldehyde, 2-methylpropanal, ethyl propanoate and ethyl vanillin were found to have OAVs over

1. While these other compounds may not have a direct impact on the aroma of rum, they may have

an impact on flavor release based on their interactions with other compounds present in the matrix.

The concentrations of all compounds are of similar magnitude to those reported in other rum

studies (Franitza et al., 2016a, 2016b; Pino et al., 2012). Significant variation does exists with respect

to concentrations between studies, but as demonstrated in this study, these differences are most

likely attributed to variations among products. OAVs were also found to be of similar magnitude to

previous studies. Pino found β-damascenone to have the highest OAV in the rum sample analyzed,

followed by ethyl butanoate, vanillin, ethyl hexanoate, guaiacol, cis-whiskey lactone, ethyl 2-

methylpropanoate, ethyl octanoate, ethyl decanoate, 4-propyl guaiacol, eugenol, acetal, ethyl 2-

methylbutanoate, γ-nonalactone, 4-ethylguaiacol, 2-phenethyl alcohol, isoamyl acetate and 2-

phenethyl acetate, which all had an OAVs greater than 1 (Pino et al., 2012). These results are similar

to the present study, although the OAVs of β-damascenone and ethyl butanoate reported in the

previous study were much higher than any of the rums analyzed in this present study.

Franitza’s analysis of two rum samples found only 14 compounds in rum A and 12 compounds in

rum B to have OAVs above 1 of the 37 compounds quantitated (Franitza et al., 2016a). Rum A,

their aged rum sample, contained vanillin, 2-methylbutanoate, β-damascenone, 3-methylbutanal, 2,3-

butandione, ethyl butanoate, acetal, and cis-whiskey lactone at OAVs well over 1 (OAVs ≥ 5). These

compounds were also found to be the most important in the aged rums in the present study. Rum B,

the lower quality rum, contained 2,3-butandione, 3-methylbutanal, ethyl butanoate, ethyl 2-

methylbutanote, ethyl pentanoate, β-damascenone, and acetal at high OAVs. Franitza found the

OAVs to be much higher for 3-methylbutanal and ethyl butanoate in the lower quality rum

compared to the aged rum, while in our study the aged rums tended to have higher concentrations

of all compounds. The lower OAVs in Rum A could be due to the fact that the rum was aged in a

84

solera system, as we also observed lower compound concentrations and OAVs for DX solera aged

rum as well.

Franitza’s (2016b) more recent study following the production process of rum. They observed an

increase in concentration and OAVs for many of the compounds. This difference might be

attributed to different aging practices as in that study - the rum was single cask aged for 3 years in

contrast to the rum aged by the solera system in their previous study. β-Damascenone was found to

have the highest OAV (3280), followed by 3-methylbutanal, 2,3-butandione, ethyl 2-

methylbutaonate, vanillin, ethyl butanoate, 2-methylbutanal, guaiacol, acetal, sotolon, ethyl

octanoate, 2-methyl-1-butanol, ethyl 3-methylbutanoate, 4-ethyguaiacol, 3-methyl-1-butanol, ethyl

hexanoate, 4-propylphenol, isoamyl acetate, 4-ethylphenol and 3-methyl butyric acid, which were all

found to have OAVs of 5 or greater. The present study also found many of these compounds to

have OAV’s above 1 with β-damascenone, 3-methylbutanal, ethyl 2-methylbutanoate, vanillin, and

ethyl butanoate also having relatively high OAV’s in the present study.

Chemometrics

Although pre-processing can be a useful tool in chemometrics, only exclusion of certain variables

was applied, and no data transformation was employed other than what was done by SAS.

Preliminary analyses were done on unmanipulated and autocorrected (application of mean centering

followed by variance scaling (dividing by standard deviation)), revealed no significant differences in

the data analysis.

Four different PCA analyses were performed to correlate the aroma sensory attributes obtained

from a previous descriptive analysis panel study (Chapter 6, Table 6.3) to various analytical

measurements, specifically: 1) quantitation data, 2) odor activity values, 3) flavor dilution factors of 8

or greater, and 4) flavor dilution factors of 16 or greater. These different analytical measurements

were selected to evaluate which measurement may be most appropriate for correlating the analytical

and sensory data.

Aroma intensity ratings and quantitation data was correlated and the results can be seen in Figure

4.4. Pearson correlation coefficients used to create the PCA plot are given in Table 4.5. The first

principal component (PC1) accounted for 44.6% of the variation, with the second principal

component (PC2) accounting for 23.0% of the variation. Visually apparent in the graph, the aroma

85

attributes are clustered into two distinct groups with roasted aroma in the middle. One group

contains vanilla, caramel, chocolate, brown sugar, coconut and maple aromas and the other contains

smoky, phenolic, alcohol and citrus aromas. Many of the odor-active compounds quantitated were

not significantly correlated to any of the aroma attributes. As expected, ethyl vanillin was highly

correlated (p<0.01) to the sweet aroma attributes including brown sugar, caramel, maple, vanilla and

chocolate aroma while being negatively correlated to alcohol, phenolic and citrus aroma. β-

Damascenone was also significantly correlated (p<0.1) to brown sugar aroma and maple aroma.

Ethyl vanillin has a sweet, vanilla odor and it is logical that the high levels present in DX and DR12

would increase perceptions of those sweet-like aroma attributes. Additionally, ethyl butyrate was

significantly correlated to maple aroma and negatively correlated to both citrus and alcohol aroma.

2-Methylpropanal and ethyl propanoate were significantly correlated to phenolic aroma, which is

interesting as the compounds have chocolate and fruity aromas, respectively, when perceived by

alone. Additionally, acetic acid was correlated to coconut aroma. Finally roasted aroma has

significant negative correlation with acetal, ethyl propanoate, ethyl 2-methylbutanoate, ethyl 3-

methylbutanoate, ethyl pentanoate, trans-whiskey lactone, cis-whiskey lactone, (E)-isoeugenol, ethyl

vanillate, and syringaldehyde.

Aroma data was also correlated with OAV data. The PCA results can be seen in Figure 4.5 and the

Pearson correlation coefficients are provided in Table 4.6. The first principal component (PC1)

accounted for 38.9% of the variation, with the second principal component (PC2) accounting for

26.6% of the variation. Compounds that had OAVs of less than 1 for all nine rums were not used

for the analysis as those compounds are not expected to significantly impact the overall aroma

perception of the rum. Similar to the correlation with quantitation data, roasted aroma was

negatively correlated (p<0.1) to acetal, ethyl propanoate, ethyl 2-methylbutanoate, ethyl 3-

methylbutanoate, 2-methyl-1-butanol, 3-methyl-1-butanol, ethyl octanoate, 2-phenethyl alcohol, 4-

ethylguaiacol and eugenol. Additionally, phenolic aroma was correlated to methylpropanal and 2-

methylbutanal. β-Damascenone was again correlated to brown sugar and maple aroma as well as

coconut aroma while negatively correlated to citrus aroma. Interestingly, ethyl vanillin had no

significant correlation with any of the aroma attribute when the results were converted to OAVs,

however, vanillin had a number of significant correlations. Vanillin was correlated to brown sugar,

caramel, vanilla, roasted and chocolate aroma and negatively correlated to phenolic aroma. Finally,

86

ethyl hexanoate was correlated to brown sugar, caramel, and maple aroma, while negatively

correlated to alcoholic aroma.

Finally, the aroma attributes were correlated to flavor dilution factors. Two different levels of FD

factors were chosen as cut off for evaluation to see how changes in selected variables may affect the

outcome of the correlation. The first cutoff was that compounds had to be perceived at an FD

factor of 8 or greater in at least one of the rum samples. Any compounds that did not meet this

criterion were removed from the analysis. The PCA results can be seen in Figure 4.6 and the

Pearson correlation coefficients are provided in Table 4.7. The first principal component (PC1)

accounted for 27.6% of the variation, with the second principal component (PC2) accounting for

19.8% of the variation. Similar to the correlations with the quantitation data, ethyl vanillin was highly

correlated to brown sugar, caramel, maple, vanilla and chocolate aroma while negatively correlation

to citrus, phenolic, alcohol and smoky aroma. However, unlike the quantitation and OAV

correlations roasted aroma did not have as many significant correlations. Roasted aroma was

correlated to ethyl butanoate and syringaldehyde while negatively correlated to eugenol and (Z)-6-

dodecen-γ-lactone. In contrast, smoky aroma was significantly correlated with a number of

compounds, where no significant correlation were observed in the first two PCA plots. Smoky

aroma was positively correlated to 2-/3-methylbuatnal, ethyl 2-methylpropanoate, ethyl-2-

methylbutanoate, ethyl cyclohexanoate, p-cresol, 4-propylguaiacol, and sotolon, while negatively

correlated to ethyl vanillin. Ethyl butanoate was another compound with numerous correlations

including positive correlations with caramel, vanilla, roasted and chocolate aromas and a negative

correlation with phenolic aroma.

Aroma attributes were then correlated to compounds with FD factors of 16 or greater in at least one

of the rum samples. The PCA results can be seen in Figure 4.7, and the Pearson correlation

coefficients are provided in Table 4.8. The first principal component (PC1) accounted for 29.3% of

the variation, with the second principal component (PC2) accounting for 21.3% of the variation.

Moving the cutoff to an FD factor of 16 or greater in at least one of the rums eliminated

acetaldehyde, ethyl 2-methylpropanoate, 4-ethylguaiacol, 4-propyl guaiacol, isoeugenol and (Z)-6-

dodecen-γ-lactone from the analysis. Interestingly, the removal of these compounds caused no

change in the significant correlation or even the correlation coefficients other than the removal of

those linked to compounds omitted from the analysis. The results suggest that the cutoff for FD

factors to consider in correlation to sensory results does not have a huge impact on the overall

87

correlations other than to remove possible correlations that may exist between those lower

perceived compounds and the aroma attributes being evaluated.

Comparing the four PCAs to each other, the correlations using the quantitation and OAV data

accounted for the most variation amongst the samples (68.6% and 65.5% respectively). The

correlations with FD factors only account for about half of the variation (47.7% for FD ≥ 8 and

50.6% for FD ≥ 16). All four analyses had the aroma attributes clustered into two distinct groups,

brown sugar, caramel, chocolate, coconut, maple and vanilla aromas in one group and alcohol,

citrus, phenolic and smoky aromas in the other. Of interest is that fact that ethyl vanillin was highly

correlated with brown sugar, caramel, chocolate, maple and vanilla aromas in all PCA except the

correlation with OAVs. In that PCA plot, vanillin was significantly correlated to those attributes, and

ethyl vanillin had no significant correlations. The other main difference was that the quantitation and

OAV PCA plots saw significant correlations with roasted aroma, while for the FD factors PCA

showed significant correlations with smoky aroma. Interestingly, roasted aroma seems to be better

defined by the lack of certain compounds rather than the presence of others as indicated by the

significant number of negative correlations.

While results show that the the analytical variables selected to correlate with sensory attributes has

an impact on the final results, all four PCA’s saw similar correlations. Based on the present study,

OAV and quantitation data should be further pursued when correlating sensory and analytical results

as both PCAs accounted for over 65% of the data variation.

Overall, the number of correlations are not as significant or insightful as may have been initially

thought. The lack of significant correlations suggests that the changes in sensory perceptions are

more complex, determined by changes in multiple variables. While flavors may be affected by

changes in a single compound, the reality of the situation is that the concentrations of all

compounds differ simultaneously relative to each other from rum to rum. Therefore, changes in

sensory perception that may be caused my synergies between compounds are difficult to detect and

understand, requiring a significant amount of work to tease out the relationships of how compounds

interact together.

Additionally, evaluation of a greater number and variety of rums may provide greater insights into

how changes in volatile composition change sensory perception. Nine rums is a relatively small

number of samples, much smaller than the number of variables evaluated. Increasing the number of

88

samples may increase the power of the chemometric analysis to better tease out relationships

between sensory perceptions and volatile concentrations.

89

4.5 Tables and Figures

Table 4.1 Compounds, isotopes, selected ions, and response factors used for stable isotope dilution analysis

Selected Ion (m/z)a

No.b Target Isotope Unlabeled Labeled Rf c

1 acetaldehyde -d - 1.253

2 2-methylpropanal d2-2-methylpropanal 72 74 0.579

4 acetal d10-acetal 103 113 0.819

5a 3-methylbutanal d2-3-methylbutanal 71 73 0.488

5b 2-methylbutanal d2-2-methylbutanal 86 88 0.518

6 ethyl propanoate d5-ethyl propanoate 57 62 1.136

7 ethyl 2-methylpropanoate d5-ethyl 2-methylpropanoate 116 121 1.220

12 ethyl butanoate d7-ethyl butanoate 71 78 0.553

13 ethyl 2-methylbutanoate d5-ethyl 2-methylbutanoate 102 107 0.980

15 ethyl 3-methylbutanoate d2-ethyl 3-methylbutanoate 115 117 0.122

17 isobutanol d2-isobutanol 74 76 0.656

18 isoamyl acetate d2-isoamyl acetate 70 72 0.166

19 ethyl pentanoate d5-ethyl pentanoate 88 93 0.812

21a 3-methyl-1-butanol d2-3-methyl-1-butanol 70 72 0.711

21b 2-methyl-1-butanol d2-2-methyl-1-butanol 70 72 0.478

22 ethyl hexanoate d11-ethyl hexanoate 99 110 2.018

25 ethyl octanoate d4-ethyl octanoate 127 131 0.580

26 acetic acid -d - 0.743

34 β-damascenone d4-β-damascenone 69 73 0.261

35 guaiacol d3-guiacol 124 127 1.384

36a trans-whiskey lactone d2-trans-whiskey lactone 99 101 0.342

37 2-phenethyl alcohol 13C2-2-phenethyl alcohol 91 93 1.177

38a cis-whiskey lactone d2-cis-whiskey lactone 99 101 1.366

38b 4-methylguaiacol d3-4-methylguaiacol 123 125 0.393

41 4-ethylguaiacol d5-4-ethylguaiacol 152 157 0.763

43 p-cresol d3-p-cresol 108 111 0.704

44 m-cresol d8-m-cresol 108 115 0.446

46 eugenol d3-eugenol 164 167 0.760

48 syringol d3-syringol 154 157 0.789

49 (E)-isoeugenol d3-(E)-isoeugenol 164 167 0.283

53 ethyl vanillin d3-vanillin 166 155 0.416

54 vanillin d3-vanillin 152 155 0.229

56 ethyl vanillate d5-ethyl vanillate 196 201 0.380

58 syringaldehyde d3-syringaldehyde 182 185 0.603

aIons used for quantitation. bNumbers correspond to identification of compounds in Table 3.2. cRespose factors determined by analyzing the area ratios of target and for a variety of mass ratios (5:1, 2:1, 1:1, 1:2, 1:5). dCompound quantitated by direct injection using external calibration.

90

Figure 4.1 Isotopically labeled standards used for quantitation

d7-ethyl butanoate d5-ethyl 2-methylbutanoate

d2-ethyl 3-methylbutanoate d2-isobutanol

d2-methylpropanal d10-acetal d2-3-methylbutanal

d2-2-methylbutanal

d5-ethyl propanoate

d2-isoamyl acetate

d5-ethyl 2-methylpropanoate

d5-ethyl pentanoate

91

Figure 4.1 (cont.) Isotopically labeled standards used for quantitation

d2-3-methyl-1-butanol

(pentanoate

d2-2-methyl-1-butanol

pentanoate

d2.-trans-whiskey lactone

d3.-4-methylguaiacol 13C2.-2-phenethyl alcohol

d2.-cis-whiskey lactone

d11-ethyl hexanoate d4-ethyl octanoate

d4-β-damascenone d3.-guaiacol

d5.-4-ethylguaiacol

92

Figure 4.1 (cont.) Isotopically labeled standards used for quantitation

d3.-syringaldehyde

d3.-p-cresol d3.-m-cresol

d3.-eugenol

d3.-syringol

d3.-(E)-isoeugenol

d3.-vanillin

d5.-ethyl vanillate

93

Table 4.2 Concentrations of the most important aroma compounds in nine rums

Concentration ug/L (ppb) (±%RSD)†

No. Compound BW BG AE RA7 AE12 DR12 ED12 RZ DX

1 acetaldehydeα 42700 (± 12.4)f 57500 (± 11.7)e 116000 (± 1.4)b 86100 (± 6.6)c 143000 (± 9.7)a 93100 (± 8.7)c 73300 (± 8.5)d 113000 (± 5.7)b 28000 (± 9.0)g

2 2-methylpropanalβ 307 (± 5.0)d 206 (± 2.9)d,e 2360 (± 5.4)b 303 (± 0.78)d 2880 (± 5.2)a 182 (± 10.7)e 784 (± 4.6)c 104 (± 5.5)e,f 38.0 (± 10.1)f

4 acetalγ 11100 (± 5.2)h 15400 (± 3.6)f 25500 (± 2.4)e 28000 (± 2.0)d 44600 (± 1.0)a 13500 (± 1.9)g 37600 (± 1.1)b 35200 (± 1.8)c 7500 (± 1.9)i

5a 3-methylbutanalβ 126 (± 7.9)f 225 (± 0.87)e 422 (± 3.8)c 358 (± 4.3)d 1210 (± 2.5)a 369 (± 7.4)d 806 (± 6.8)b 34.7 (± 5.7)d 63.0 (± 6.9)g

5b 2-methylbutanalβ 160 (± 10.3)e 137 (± 4.8)e,f 1230 (± 6.2)b 219 (± 6.2)d 1520 (± 3.7)a 83.9 (± 4.2)f,g 487 (± 6.8)c 57.3 (± 3.1)g 39.8 (± 7.1)g

6 ethyl propanoateβ 273 (± 10.3)f 306 (± 11.7)f 1080 (± 3.5)c 793 (± 5.2)d 1890 (± 9.2)a 737 (± 9.5)d 1430 (± 9.2)b 595 (± 2.9)e 861 (±2.3)d

7 ethyl 2-methylpropanoateβ 70.5 (± 4.9)f 133 (± 3.4)d 156 (±2.7)c 117 (± 3.1)e 270 (± 1.2)b 55.0 (± 4.5)g 143 (± 1.2) c,d 692 (± 3.2)a 12.6 (± 11.3)h

12 ethyl butanoateβ 123 (± 5.8)g 208 (± 3.2)f 366 (± 5.2)d 156 (± 8.1)g 547 (± 6.5)c 279 (± 9.1)e 608 (± 6.2)b 223 (± 3.0)f 650 (± 0.41)a

13 ethyl 2-methylbutanoateβ 7.54 (± 6.2)e 17.7 (± 3.6)d 29.6 (± 7.0)b 18.9 (± 8.9)d 81.1 (± 6.6)a 8.23 (± 4.5)e 26.0 (± 3.3)c 10.3 (± 3.4)e 3.89 (± 6.5)f

15 ethyl 3-methylbutanoateβ 20.6 (± 10.3)f 60.5 (± 7.4)d,e 93.2 (± 1.2)c 70.8 (± 10.5)d 238 (± 9.8)a 45.7 (± 6.9)e 127 (± 8.7)b 64.8 (± 4.7)d 14.6 (± 7.6)f

17 isobutanolβ 84000 (± 3.9)e 104000 (± 8.9)e 177000 (± 9.4)d 209000 (± 11.6)c,d 235000 (± 5.7)b,c 212000 (± 5.5)b,c,d 494000 (± 10.0)a 247000 (± 5.7)b 31100 (± 1.8)f

18 isoamyl acetateδ 118 (± 2.8)g,h 123 (± 5.3)g 853 (± 4.3)b 189 (± 5.1)f 502 (± 2.4)e 737 (± 2.2)c 1070 (± 2.5)a 537 (± 1.6)d 88.9 (± 8.5)h

19 ethyl pentanoateβ 7.58 (± 9.3)f,g 10.5 (± 9.1)e,f 23.5 (± 9.2)c 20.9 (± 5.7)c,d 83.8 (± 11.9)a 11.0 (± 4.3)e,f 36.6 (± 6.1)b 15.6 (± 7.8)d,e 3.23 (± 10.2)g

21a 3-methyl-1-butanolβ 134000 (± 0.50)h 166000 (± 3.8)g 271000 (± 9.2)e 238000 (± 1.5)f 367000 (± 0.78)d 393000 (± 4.3)c 415000 (± 0.99)b 539000 (± 0.59)a 46400 (± 0.96)i

21b 2-methyl-1-butanolβ 14500 (± 4.1)f 17300 (± 8.5)f 36000 (± 4.1)e 30500 (± 9.0)e 56100 (± 5.1)d 211000 (± 3.1)*, a 65800 (± 11.3)c 79200 (± 4.1)b 5480 (± 9.3)g

22 ethyl hexanoateδ 246 (± 7.7)f 371 (± 5.0)d,e 524 (± 6.0)c 350 (± 2.8)e 1280 (± 2.5)b 466 (± 7.3)c,d 1580 (± 9.8)a 485 (± 2.6)c 108 (± 6.6)g

25 ethyl octanoateδ 70.9 (± 3.6)g 69.8 (± 7.2)*, g 564 (± 2.9)c 313 (± 0.82)d 818 (± 3.3)b 309 (± 1.6)d 1800 (± 2.1)a 761 (± 9.0)b 288 (± 7.0)h

26 acetic acidα 15400 (± 7.7)g 24200 (± 4.4)g 110000 (± 10.9)f 628000 (± 4.69)b 372000 (± 0.81)d 785000 (± 3.5)a 765000 (± 6.7)a 502000 (± 3.4)c 164000 (± 8.5)e

34 β-damascenoneδ 3.38 (± 2.1)h 8.99 (± 7.0)g 36.5 (± 2.0)b 27.2 (± 5.6)c 23.8 (± 0.96)d 14.3 (± 4.6)f 15.7 (± 6.67)e 9.60 (± 8.1)g 43.9 (± 1.2)a

35 guaiacolδ 89.3 (± 4.8)e 37.6 (± 5.4)f 90.3 (± 5.3)e 255.3 (± 8.4)c 347 (± 5.4)a 151 (± 2.6)d 309 (± 3.6)b 104 (± 7.4)e 8.00 (± 7.6)g

36a trans-whiskey lactoneδ 5.63 (± 5.2)h 8.91 (± 3.4)g,h 36.5 (± 4.2)e 74.9 (± 4.8)c 126 (± 3.9)b 21.7 (± 3.5)f 140.9 (± 3.5)a 60.8 (± 6.4)d 13.7 (± 2.0)g

37 2-phenethyl alcoholδ 882 (± 0.85)g 944 (± 2.5)*, g 1330 (± 1.1)e 2420 (± 0.69)d 2700 (± 2.8)c 1170 (± 0.62)f 4360 (± 1.7)a 3510 (± 2.5)b 748 (± 0.007)h

38a cis-whiskey lactoneδ 125 (± 3.9)i 268 (± 1.6)h 1270 (± 3.9)e 3020 (± 3.5)c 4920 (± 0.92)b 707 (± 4.6)f 5350 (± 2.5)a 2210 (± 3.5)d 464 (± 2.5)g

38b 4-methylguaiacoleδ 1.16 (± 11.2)g 4.33 (± 0.62)*, f 7.08 (± 7.3)c,d 1.69 (± 1.51)*, a 13.1 (± 8.3)b 5.64 (± 11.2)e 8.09 (± 3.7)c 6.79 (± 3.5)d,e 1.53 (± 3.7)g

94

Table 4.2 (cont.) Concentrations of the most important aroma compounds in nine rums

Concentration ug/L (ppb) (±%RSD)a

No. Compound BW BG AE RA7 AE12 DR12 ED12 RZ DX

41 4-ethylguaiacolδ N.D.g 1.12 (± 9.1)f 17.6 (± 7.6)b 11.0 (± 10.22)d 32.2 (± 1.1)a 17.0 (± 0.51)b 5.20 (± 1.6)e 14.4 (± 3.5)c 0.831 (± 6.1)f,g

43 p-cresolδ 1.72 (± 1.2)c 0.870 (± 0.22)*,f 1.63 (± 4.3)c 2.24 (± 6.4)*, a 1.97 (± 1.1)b 1.13 (± 6.7) e 0.902 (± 8.9)f 1.09 (± 6.98)e 1.07 (± 6.1)*, d

44 m-cresolδ 1.08 (± 8.82)d 1.21 (± 6.4)d 1.08 (± 3.7)d 2.54 (± 3.7)b 3.57 (± 1.6)a 1.00 (± 4.2)d 3.77 (± 8.7)a 1.68 (± 2.02)c 0.402 (± 9.2),e

46 eugenolδ ND 31.6 (± 4.3)c 31.4 (± 5.0)c 39.8 (± 1.8)b 24.0 (± 3.3)d 14.2 (± 6.6)f 94.0 (± 1.4)a 24.4 (± 3.8)d 16.9 (± 2.1)e

48 syringolδ 6.44 (± 0.24)e,f 4.79 (± 8.3)f 9.87 (± 7.5)d 16.30 (± 4.5)c 43.50 (± 6.2)a 9.19 (± 5.7)d 30.30 (± 3.9)b 9.90 (± 7.8)d 6.74 (± 5.9)e

49 (E)-isoeugenolδ 2.23 (± 6.1)*,e 118 (± 10.6)c 99.2 (± 8.9)c 46.7 (± 8.1)d 678 (± 7.8)a 36.4 (± 5.0)d,e 385 (± 6.1)b 46.7 (± 8.06)d 7.84 (± 10.3)d,e

53 ethyl vanillinε N.D. N.D. N.D. N.D. N.D. 873 (± 6.4) N.D. N.D. 3020 (± 3.4)

54 vanillinδ 322 (± 1.7)g 325 (± 1.7)g 1180 (± 0.50)f 2270 (± 1.0)d 4470 (± 2.3)b 3050 (± 4.1)c 4780 (± 3.4)a 1940 (± 0.26)e 1200 (± 5.3)f

56 ethyl vanillateδ 111 (± 1.1)g 126 (± 0.93)f,g 271 (± 0.71)d 582 (± 4.8)c 1430 (± 6.1)a 177 (± 5.0)e,f 1080 (± 1.6)b 218 (± 1.3)d,e 40.3 (± 6.6)h

58 syringaldehydeδ 713 (± 2.8)h 1040 (± 4.8)*,g 2020 (± 2.4)e 3880 (± 3.8)c 8580 (± 1.6)a 1590 (± 0.22)f 6910 (± 1.6)b 2970 (± 3.08)*,d 461 (± 1.2)i

†Mean values of triplicates. αCalculated using external standard methodology and direct injection. βCalculated using HS-SPME in combination with SIDA. γCalculated by direct injection in combination with SIDA. δCalculated using extraction and liquid injection in combination with SIDA. εCalculated using extraction in combination with internal standard methodology. +N.D., not detected *Mean values of duplicate rather than triplicate. a,b,c,d,e,f,g,h,i indicate values statistically different across rums. ‡ Chirality determined in previous paper: (S)-Ethyl 2-methylbutanoate > 99%, (S)-2-methylbutanol >99% (Franitza et al., 2016a) ^No difference detected.

95

Table 4.3 ANOVA F-ratios for quantitation data

No. Compound Rum Pr

1 acetaldehyde 82.59 <0.0001

2 2-methylpropanal 750.34 <0.0001

4 acetal 2066.08 <0.0001

5a 3-methylbutanal 1011 <0.0001

5b 2-methylbutanal 777.27 <0.0001

6 ethyl propanoate 124.78 <0.0001

7 ethyl 2-methylpropanoate 1942.4 <0.0001

12 ethyl butanoate 275 <0.0001

13 ethyl 2-methylbutanoate 403.2 <0.0001

15 ethyl 3-methylbutanoate 163 <0.0001

17 isobutanol 122.41 <0.0001

18 isoamyl acetate 1322 <0.0001

19 ethyl pentanoate 144.21 <0.0001

21a 3-methyl-1-butanol 662.55 <0.0001

21b 2-methyl-1-butanol 483 <0.0001

22 ethyl hexanoate 233 <0.0001

25 ethyl octanoate 854 <0.0001

26 acetic acid 528 <0.0001

34 β-damascenone 866 <0.0001

35 guaiacol 372 <0.0001

36a trans-whiskey lactone 870 <0.0001

37 2-phenethyl alcohol 1989 <0.0001

38a cis-whiskey lactone 2656 <0.0001

38b 4-methylguaiacol 219 <0.0001

41 4-ethylguaiacol 1037 <0.0001

43 p-cresol 117 <0.0001

44 m-cresol 218 <0.0001

46 eugenol 2075 <0.0001

48 syringol 429 <0.0001

49 (E)-isoeugenol 326 <0.0001

54 vanillin 1268 <0.0001

56 ethyl vanillate 734 <0.0001

58 syringaldehyde 3145 <0.0001

96

Figure 4.2 Cluster analysis of nine rums based on quantitation data

97

Table 4.4 Odor activity values and odor thresholds for most important compounds in rum

OAVa

No. Compound Threshold (ug/L) BW BG AE RA7 AE12 DR12 ED12 RZ DX

1 acetaldehyde 192000b 0.2 0.3 1 0.4 1 0.5 0.4 1 0.1

2 2-methylpropanal 5.9c 52 35 400 51 488 31 133 18 6

4 acetal 719d 15 21 35 39 62 19 52 49 10

5a 3-methylbutanal 2.8d 45 80 151 128 432 132 288 12 23

5b 2-methylbutanal 33e 5 4 37 7 46 3 15 2 1

6 ethyl propanoate 3452d 0.1 0.1 0.3 0.2 1 0.2 0.4 0.2 0.2

7 ethyl 2-methylpropanoate 4.5d 16 30 35 26 60 12 32 154 3

12 ethyl butanoate 9.5d 13 22 39 16 58 29 64 23 68

13 ethyl 2-methylbutanoate 0.22d 34 80 135 86 369 37 118 47 18

15 ethyl 3-methylbutanoate 1.6d 13 38 58 44 149 29 79 41 9

17 isobutanol 160000e 1 1 1 1 1 1 3 2 0.2

18 isoamyl acetate 245d 0.5 1 3 1 2 3 4 2 0.4

19 ethyl pentanoate 11e 1 1 2 2 8 1 3 1 0.3

21a 3-methyl-1-butanol 56100d 2 3 5 4 7 7 7 10 1

21b 2-methyl-1-butanol 6100e 2 3 6 5 9 35 11 13 1

22 ethyl hexanoate 30d 8 12 17 12 43 16 53 16 4

25 ethyl octanoate 147d 0.5 0.5 4 2 6 2 12 5 2

26 acetic acid 230000e 0.1 0.1 0.5 3 2 3 3 2 1

34 β-damascenone 0.1d 34 90 365 272 238 143 157 96 439

35 guaiacol 9.2d 10 4 10 28 38 16 34 11 1

36a trans-whiskey lactone 790f <0.1 <0.1 <0.1 0.1 0.2 <0.1 0.2 0.1 <0.1

37 2-phenethyl alcohol 2600d 0.3 0.4 1 1 1 0.5 2 1 0.3

38a cis-whiskey lactone 67f 2 4 19 45 73 11 80 33 7

38b 4-methylguaiacol 315b <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1

98

Table 4.4 (cont.) Odor activity values and odor thresholds for most important compounds in rum

OAVa

No. Compound Threshold (ug/L) BW BG AE RA7 AE12 DR12 ED12 RZ DX

41 4-ethylguaiacol 6.9d -* 0.2 3 2 5 2 1 2 0.1

43 p-cresol 81.5g <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1

44 m-cresol 680h <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1

46 eugenol 7.1d - 4 4 6 3 2 13 3 2

48 syringol 570i <0.1 <0.1 <0.1 <0.1 0.1 <0.1 0.1 <0.1 <0.1

49 (E)-isoeugenol 1860j <0.1 0.1 0.1 <0.1 0.4 <0.1 0.2 <0.1 <0.1

53 ethyl vanillin 100h - - - - - 9 - - 30

54 vanillin 22d 15 15 54 103 203 139 217 88 55

56 ethyl vanillate 796000j <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1

58 syringaldehyde 6490000j <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 aOdor activity values were calculated by dividing the concentrations (Table 4.2) by the respective odor threshold. bWillner et al. 2013 (40% v/v ethanol/water matrix). c Uselmann and Schieberle 2015 (40% v/v ethanol/water matrix). dPoisson and Schieberle 2008 (40% v/v ethanol/water matrix). eFranitza et al., 2016b (40% v/v ethanol/water matrix). fOtsuka, Zenibayashi, Itoh, & Totsuka, 1974 (30% v/v ethanol/water matrix). gFranitza et al., 2016a (40% v/v ethanol/water matrix). hFazzalari, 1978 (water). iLópez, Aznar, Cacho, & Ferreira, 2002 (10% v/v ethanol/water matrix). jOdor thresholds were determined in 40% v/v ethanol/water matrix. * - indicates compound not quantifed

99

Figure 4.3 Cluster analysis of nine rums based on OAV data

100

Figure 4.4 Principal component analysis of quantitation and sensory data a) sensory and analytical variables used to create the model b) rum samples used to create model

101

Figure 4.5 Principal component analysis of odor activity values and sensory data a) sensory and analytical variables used to create the model b) rum samples used to create model

102

Figure 4.6 Principal component analysis of flavor dilution factors ≥ 8 and sensory data a) sensory and analytical variables used to create the model b) rum samples used to create model

103

Figure 4.7 Principal component analysis of flavor dilution factors ≥ 16 and sensory data a) sensory and analytical variables used to create the model b) rum samples used to create model

104

Table 4.5 Pearson correlation coefficients for quantitation and sensory (aroma) data

Attributes Brown Sugar

Aroma Caramel Aroma

Maple Aroma

Vanilla Aroma

Coconut Aroma

Citrus Aroma

Phenolic Aroma

Alcohol Aroma

Smoky Aroma

Roasted Aroma

Chocolate Aroma

acetaldehyde -0.162 -0.350 -0.328 -0.318 -0.388 0.043 0.532 0.280 0.248 -0.331 -0.285

2-methylpropanal -0.263 -0.446 -0.277 -0.447 -0.582 -0.067 0.668** 0.284 0.488 -0.530 -0.400

acetal -0.233 -0.500 -0.325 -0.456 -0.160 -0.002 0.538 0.314 0.191 -0.692** -0.344

3-methylbutanal 0.320 0.392 0.215 0.488 0.388 -0.363 -0.486 -0.335 -0.409 0.519 0.462

2-methylbutanal -0.262 -0.458 -0.275 -0.454 -0.561 -0.079 0.668** 0.290 0.494 -0.549 -0.406

ethyl propanoate 0.192 -0.094 0.145 -0.050 0.005 -0.544 0.231 -0.142 0.061 -0.605* 0.064

ethyl 2-methylpropanoate

-0.244 -0.248 -0.272 -0.296 -0.148 0.376 0.304 0.183 0.002 -0.347 -0.210

ethyl butanoate 0.496 0.360 0.593* 0.362 0.368 -0.732** -0.266 -0.596* -0.122 -0.366 0.446

ethyl 2-methylbutanoate -0.184 -0.426 -0.239 -0.432 -0.506 -0.051 0.576 0.299 0.188 -0.715** -0.288


methylpropanol -0.105 -0.308 -0.221 -0.176 0.281 -0.205 0.190 0.133 0.081 -0.257 -0.150

isoamyl acetate -0.048 -0.157 -0.148 -0.035 0.148 -0.292 0.187 -0.016 0.281 0.035 -0.101

ethyl pentanoate -0.135 -0.400 -0.207 -0.388 -0.369 -0.134 0.520 0.235 0.069 -0.759** -0.214

3-methyl-1-butanol -0.100 -0.202 -0.271 -0.131 0.058 0.030 0.218 0.116 -0.060 -0.141 -0.084

2-methyl-1-butanol 0.323 0.334 0.095 0.426 0.253 -0.273 -0.305 -0.257 -0.397 0.477 0.403

ethyl hexanoate -0.159 -0.385 -0.221 -0.302 -0.012 -0.209 0.352 0.172 0.091 -0.550 -0.189

ethyl octanoate -0.025 -0.216 -0.042 -0.126 0.305 -0.332 0.156 -0.049 0.091 -0.432 -0.062

acetic acid 0.403 0.212 0.183 0.348 0.602* -0.490 -0.302 -0.259 -0.386 0.090 0.365

β-damascenone 0.595* 0.456 0.667** 0.444 0.292 -0.690** -0.274 -0.564 0.182 -0.044 0.363

guaiacol -0.057 -0.373 -0.214 -0.276 0.001 -0.261 0.342 0.225 -0.071 -0.565 -0.141

trans-whiskey lactone -0.034 -0.339 -0.115 -0.262 0.115 -0.290 0.311 0.116 0.022 -0.674** -0.132

2-phenethyl alcohol -0.110 -0.325 -0.184 -0.250 0.254 -0.103 0.231 0.124 -0.012 -0.515 -0.152

cis-whiskey lactone -0.026 -0.339 -0.110 -0.263 0.112 -0.292 0.310 0.121 0.016 -0.683** -0.131

4-methylguaiacol -0.130 -0.302 -0.215 -0.280 -0.321 -0.078 0.445 0.154 0.145 -0.501 -0.175

4-ethylguaiacol 0.047 -0.151 -0.117 -0.135 -0.347 -0.163 0.383 0.081 0.048 -0.350 -0.055

105

Table 4.5 (cont.) Pearson correlation coefficients for quantitation and sensory (aroma) data


Aroma Caramel Aroma

Maple Aroma

Vanilla Aroma

Coconut Aroma

Citrus Aroma

Phenolic Aroma

Alcohol Aroma

Smoky Aroma

Roasted Aroma

Chocolate Aroma

p-cresol -0.196 -0.391 -0.272 -0.390 -0.403 0.033 0.494 0.359 0.155 -0.396 -0.325

m-cresol -0.183 -0.482 -0.281 -0.403 -0.013 -0.119 0.404 0.311 0.032 -0.680 -0.264

eugenol -0.024 -0.225 -0.050 -0.104 0.406 -0.253 0.050 0.077 0.279 -0.196 -0.160

syringol -0.024 -0.318 -0.099 -0.276 -0.115 -0.294 0.367 0.116 -0.061 -0.756 -0.089

(E)-isoeugenol -0.121 -0.359 -0.162 -0.342 -0.284 -0.153 0.427 0.189 0.033 -0.735** -0.166

ethyl vanillin 0.816*** 0.897*** 0.933*** 0.837*** 0.577 -0.666* -0.815*** -0.912*** -0.485 0.238 0.848***

vanillin 0.245 -0.024 0.101 0.074 0.275 -0.525 0.026 -0.154 -0.248 -0.409 0.202

ethyl vanillate -0.083 -0.389 -0.165 -0.333 -0.114 -0.247 0.406 0.199 0.015 -0.737** -0.170

syringaldehyde -0.080 -0.384 -0.175 -0.326 -0.071 -0.222 0.398 0.188 0.001 -0.731** -0.165

*, **, *** stand for significance at p<0.1, p<0.05, and p<0.01 respectively.

106

Table 4.6 Pearson correlation coefficients for odor activity values and sensory (aroma) data


Aroma Caramel Aroma

Maple Aroma

Vanilla Aroma

Coconut Aroma

Citrus Aroma

Phenolic Aroma

Alcohol Aroma

Smoky Aroma

Roasted Aroma

Chocolate Aroma

acetaldehyde -0.140 -0.254 -0.094 -0.316 -0.338 0.045 0.486 0.076 0.387 -0.555 -0.268

2-methylpropanal -0.251 -0.434 -0.247 -0.442 -0.570 -0.079 0.658* 0.262 0.491 -0.559 -0.391

acetal -0.181 -0.450 -0.215 -0.432 -0.114 -0.050 0.496 0.232 0.195 -0.791** -0.304

3-methylbutanal 0.320 0.393 0.215 0.488 0.388 -0.362 -0.487 -0.335 -0.409 0.520 0.462

2-methylbutanal -0.250 -0.450 -0.253 -0.450 -0.548 -0.092 0.661* 0.273 0.493 -0.575 -0.398

ethyl propanoate 0.016 -0.229 -0.035 -0.235 -0.307 -0.267 0.395 0.050 -0.029 -0.735** -0.047

ethyl 2-methylpropanoate -0.228 -0.232 -0.231 -0.290 -0.133 0.360 0.291 0.153 0.010 -0.389 -0.199

ethyl butanoate 0.312 0.183 0.181 0.281 0.200 -0.557 -0.116 -0.292 -0.154 0.016 0.311


ethyl 3-methylbutanoate -0.124 -0.390 -0.145 -0.392 -0.328 -0.136 0.505 0.204 0.177 -0.807*** -0.240

isoamyl acetate -0.230 -0.320 -0.153 -0.275 0.300 0.000 0.163 0.092 0.137 -0.435 -0.219

ethyl pentanoate 0.160 0.037 0.293 0.059 0.332 -0.461 0.002 -0.318 0.307 -0.356 0.046

3-methyl-1-butanol -0.113 -0.372 -0.145 -0.384 -0.387 -0.147 0.501 0.202 0.053 -0.821*** -0.192

2-methyl-1-butanol 0.125 0.016 0.218 -0.036 0.201 -0.166 0.062 -0.245 -0.012 -0.601* 0.077

ethyl hexanoate 0.639* 0.637* 0.818*** 0.557 0.537 -0.580 -0.556 -0.787** -0.315 -0.205 0.627

ethyl octanoate -0.073 -0.304 -0.047 -0.259 0.064 -0.292 0.283 0.041 0.092 -0.702** -0.123

acetic acid -0.032 -0.228 -0.046 -0.143 0.269 -0.334 0.183 -0.042 0.103 -0.463 -0.072

β-damascenone 0.608* 0.364 0.593* 0.404 0.724** -0.697** -0.407 -0.536 -0.368 -0.368 0.494

guaiacol 0.388 0.258 0.187 0.362 0.109 -0.489 -0.112 -0.213 0.130 0.418 0.216

2-phenethyl alcohol 0.044 -0.277 0.032 -0.241 0.093 -0.366 0.264 0.047 -0.037 -0.812*** -0.071

cis-whiskey lactone -0.102 -0.346 -0.118 -0.250 0.270 -0.257 0.266 0.097 0.289 -0.472 -0.234

4-ethylguaiacol -0.012 -0.325 -0.080 -0.255 0.127 -0.307 0.297 0.097 0.018 -0.708** -0.120

eugenol 0.094 -0.141 0.095 -0.185 -0.283 -0.256 0.386 -0.035 0.194 -0.683** -0.079

107

Table 4.6 (cont.) Pearson correlation coefficients for odor activity values and sensory (aroma) data


Aroma Caramel Aroma

Maple Aroma

Vanilla Aroma

Coconut Aroma

Citrus Aroma

Phenolic Aroma

Alcohol Aroma

Smoky Aroma

Roasted Aroma

Chocolate Aroma

ethyl vanillin -0.017 -0.227 -0.055 -0.098 0.425 -0.263 0.041 0.085 0.260 -0.180 -0.156

vanillin 0.615* 0.704** 0.468 0.757** 0.398 -0.470 -0.657* -0.571 -0.535 0.687** 0.705**


108

Table 4.7 Pearson correlation coefficients for FD factors ≥ 8 and sensory (aroma) data


Aroma Caramel Aroma

Maple Aroma

Vanilla Aroma

Coconut Aroma

Citrus Aroma

Phenolic Aroma

Alcohol Aroma

Smoky Aroma

Roasted Aroma

Chocolate Aroma

acetaldehyde 0.300 0.013 0.167 0.124 0.453 -0.404 -0.124 -0.020 0.058 -0.053 0.051

acetal 0.114 0.098 -0.038 0.187 -0.094 -0.228 0.034 -0.017 0.306 0.576 -0.007

2-/3-methylbutanal -0.264 -0.294 -0.254 -0.277 -0.283 0.092 0.472 0.182 0.752** 0.004 -0.432

ethyl 2-methylpropanoate -0.094 -0.383 -0.162 -0.344 -0.282 -0.094 0.484 0.326 0.590 -0.395 -0.408

ethyl butanoate 0.575 0.646* 0.401 0.713** 0.366 -0.419 -0.619* -0.482 -0.484 0.715** 0.640*

ethyl 2-methylbutanoate -0.181 -0.400 -0.196 -0.331 -0.211 -0.206 0.520 0.235 0.752** -0.242 -0.447

unknown14 -0.377 -0.313 -0.378 -0.310 -0.306 0.581 0.163 0.533 0.427 0.218 -0.450

2-/3-methyl-1-butanol 0.176 0.284 0.030 0.313 0.155 0.009 -0.191 -0.211 -0.140 0.510 0.226

ethyl cyclohexanecarboxylate

-0.213 -0.229 -0.208 -0.183 -0.231 -0.008 0.382 0.157 0.806*** 0.230 -0.404

(Z)-2-nonenal 0.158 -0.172 0.019 -0.088 0.268 -0.280 0.091 0.116 -0.021 -0.385 -0.053

3-methylbutyric acid -0.123 -0.077 -0.101 -0.120 -0.016 0.227 0.186 -0.004 0.220 -0.078 -0.155

unknown32 -0.237 -0.183 -0.224 -0.216 -0.065 0.442 0.195 0.198 0.346 -0.005 -0.302

β-damascenone -0.209 -0.099 -0.283 -0.122 -0.224 0.561 0.098 0.290 0.237 0.293 -0.240

guaiacol -0.173 -0.129 -0.178 -0.177 -0.048 0.320 0.201 0.066 0.058 -0.172 -0.149

trans-whiskey lactone 0.080 0.010 0.016 0.146 0.486 -0.326 -0.202 -0.106 -0.084 0.127 0.121

2-phenethyl alcohol -0.020 0.053 -0.070 0.025 0.089 0.232 0.010 -0.061 -0.062 0.063 0.017

cis-whiskey lactone 0.201 0.259 0.279 0.203 0.361 -0.041 -0.209 -0.370 -0.115 -0.094 0.220

unknown40 0.356 0.393 0.519 0.357 0.191 -0.451 -0.217 -0.507 0.358 0.162 0.218

4-ethylguaiacol 0.487 0.298 0.412 0.282 -0.064 -0.585* -0.075 -0.374 -0.309 -0.347 0.422

p-cresol -0.156 -0.163 -0.176 -0.129 -0.329 -0.037 0.370 0.120 0.721** 0.254 -0.328

m-cresol -0.065 -0.025 -0.125 -0.040 0.001 0.212 0.132 0.009 0.146 0.097 -0.102

4-propylguaiacol -0.458 -0.536 -0.459 -0.520 -0.423 0.439 0.542 0.576 0.839*** -0.098 -0.683**

eugenol 0.057 -0.190 0.034 -0.230 -0.312 -0.205 0.368 0.037 -0.012 -0.794** -0.044

sotolon -0.052 -0.146 -0.170 -0.060 -0.150 -0.081 0.274 0.170 0.687** 0.349 -0.302

syringol -0.117 -0.261 -0.113 -0.143 0.322 -0.272 0.119 0.051 0.247 -0.202 -0.163

109

Table 4.7 (cont.) Pearson correlation coefficients for FD factors ≥ 8 and sensory (aroma) data


Aroma Caramel Aroma

Maple Aroma

Vanilla Aroma

Coconut Aroma

Citrus Aroma

Phenolic Aroma

Alcohol Aroma

Smoky Aroma

Roasted Aroma

Chocolate Aroma

(E)-isoeugenol 0.132 -0.021 -0.007 -0.030 -0.377 -0.151 0.189 0.040 -0.216 -0.334 0.111

(Z)-6-dodecenelactone -0.033 -0.242 -0.095 -0.289 -0.439 -0.062 0.411 0.146 -0.149 -0.751** -0.066

unknow51 -0.117 -0.085 -0.143 -0.129 0.027 0.412 0.050 0.127 -0.035 -0.131 -0.108

ethyl vanillin 0.845*** 0.945*** 0.828*** 0.941*** 0.577 -0.671** -0.869*** -0.876*** -0.603* 0.544 0.917***

vanillin 0.279 0.282 0.083 0.300 0.160 -0.078 -0.205 -0.228 -0.514 0.155 0.372

unknown55 0.414 0.460 0.526 0.373 0.481 -0.193 -0.434 -0.531 -0.382 -0.126 0.442

ethyl vanillate 0.246 -0.049 0.047 0.000 -0.025 -0.255 0.075 0.096 -0.218 -0.328 0.089

syringaldehyde 0.244 0.247 0.083 0.315 0.036 -0.193 -0.088 -0.151 0.264 0.613* 0.102

unknown59 0.409 0.444 0.197 0.506 0.146 -0.305 -0.362 -0.320 -0.459 0.516 0.492


110

Table 4.8 Pearson correlation coefficients for FD factors ≥ 16 and sensory (aroma) data


Aroma Caramel Aroma

Maple Aroma

Vanilla Aroma

Coconut Aroma

Citrus Aroma

Phenolic Aroma

Alcohol Aroma

Smoky Aroma

Roasted Aroma

Chocolate Aroma

acetal 0.114 0.098 -0.038 0.187 -0.094 -0.228 0.034 -0.017 0.306 0.576 -0.007

2-/3-methylbutanal -0.264 -0.294 -0.254 -0.277 -0.283 0.092 0.472 0.182 0.752** 0.004 -0.432

ethyl butanoate 0.575 0.646* 0.401 0.713** 0.366 -0.419 -0.619* -0.482 -0.484 0.715** 0.640*

ethyl 2-methylbutanoate -0.181 -0.400 -0.196 -0.331 -0.211 -0.206 0.520 0.235 0.752** -0.242 -0.447

unknown14 -0.377 -0.313 -0.378 -0.310 -0.306 0.581 0.163 0.533 0.427 0.218 -0.450

2-/3-methyl-1-butanol 0.176 0.284 0.030 0.313 0.155 0.009 -0.191 -0.211 -0.140 0.510 0.226

ethyl cyclohexanecarboxylate

-0.213 -0.229 -0.208 -0.183 -0.231 -0.008 0.382 0.157 0.806*** 0.230 -0.404

(Z)-2-nonenal 0.158 -0.172 0.019 -0.088 0.268 -0.280 0.091 0.116 -0.021 -0.385 -0.053

3-methylbutyric acid -0.123 -0.077 -0.101 -0.120 -0.016 0.227 0.186 -0.004 0.220 -0.078 -0.155

unknown32 -0.237 -0.183 -0.224 -0.216 -0.065 0.442 0.195 0.198 0.346 -0.005 -0.302

β-damascenone -0.209 -0.099 -0.283 -0.122 -0.224 0.561 0.098 0.290 0.237 0.293 -0.240

guaiacol -0.173 -0.129 -0.178 -0.177 -0.048 0.320 0.201 0.066 0.058 -0.172 -0.149

trans-whiskey lactone 0.080 0.010 0.016 0.146 0.486 -0.326 -0.202 -0.106 -0.084 0.127 0.121

2-phenethyl alcohol -0.020 0.053 -0.070 0.025 0.089 0.232 0.010 -0.061 -0.062 0.063 0.017

cis-whiskey lactone 0.201 0.259 0.279 0.203 0.361 -0.041 -0.209 -0.370 -0.115 -0.094 0.220

unknown40 0.356 0.393 0.519 0.357 0.191 -0.451 -0.217 -0.507 0.358 0.162 0.218

p-cresol -0.156 -0.163 -0.176 -0.129 -0.329 -0.037 0.370 0.120 0.721** 0.254 -0.328

m-cresol -0.065 -0.025 -0.125 -0.040 0.001 0.212 0.132 0.009 0.146 0.097 -0.102

eugenol 0.057 -0.190 0.034 -0.230 -0.312 -0.205 0.368 0.037 -0.012 -0.794** -0.044

sotolon -0.052 -0.146 -0.170 -0.060 -0.150 -0.081 0.274 0.170 0.687** 0.349 -0.302

syringol -0.117 -0.261 -0.113 -0.143 0.322 -0.272 0.119 0.051 0.247 -0.202 -0.163

(E)-isoeugenol 0.132 -0.021 -0.007 -0.030 -0.377 -0.151 0.189 0.040 -0.216 -0.334 0.111

ethyl vanillin 0.845*** 0.945*** 0.828*** 0.941*** 0.577 -0.671** -0.869*** -0.876*** -0.603* 0.544 0.917***

vanillin 0.279 0.282 0.083 0.300 0.160 -0.078 -0.205 -0.228 -0.514 0.155 0.372

unknown55 0.414 0.460 0.526 0.373 0.481 -0.193 -0.434 -0.531 -0.382 -0.126 0.442

111

Table 4.8 (cont.) Pearson correlation coefficients for FD factors ≥ 16 and sensory (aroma) data


Aroma Caramel Aroma

Maple Aroma

Vanilla Aroma

Coconut Aroma

Citrus Aroma

Phenolic Aroma

Alcohol Aroma

Smoky Aroma

Roasted Aroma

Chocolate Aroma

ethyl vanillate 0.246 -0.049 0.047 0.000 -0.025 -0.255 0.075 0.096 -0.218 -0.328 0.089

syringaldehyde 0.244 0.247 0.083 0.315 0.036 -0.193 -0.088 -0.151 0.264 0.613* 0.102

unknown59 0.409 0.444 0.197 0.506 0.146 -0.305 -0.362 -0.320 -0.459 0.516 0.492


112

4.6 References

Adkins, H., & Nissen, B. H. (1941). Acetal. Organic Syntheses, 1, 1.

https://doi.org/10.15227/orgsyn.003.0001

Andreu-Sevilla, A. J., Mena, P., Martí, N., García Viguera, C., & Carbonell-Barrachina, Á. A. (2013).

Volatile composition and descriptive sensory analysis of pomegranate juice and wine. Food

Research International, 54(1), 246–254. https://doi.org/10.1016/j.foodres.2013.07.007

Aznar, M., López, R., Cacho, J., & Ferreira, V. (2003). Prediction of Aged Red Wine Aroma

Properties from Aroma Chemical Composition . Partial Least Squares Regression Models.

Journal of Agricultural and Food Chemistry, 51, 2700–2707. https://doi.org/10.1021/jf026115z

Beebe, K. R., Pell, R. J., & Seasholtz, M. B. (1998). Chemometrics - A Practical Guide. New York: John

Wiley & Sons, Inc.

Brereton, R. G. (2007). Pattern Recognition. In Applied Chemometrics for Scientitsts (pp. 145–191). West

Sussex, England: John Wiley& Sons, Ltd.

Brereton, R. G. (2009). Exploratory Data Analysis. In Chemometrics for Pattern Recognition (pp. 47–106).

John Wiley& Sons, Ltd.




Cardoso, D. R., Andrade-Sobrinho, L. G., Leite-Neto, A., Reche, R. V., Isique, W. D., Ferreira, M.

M. C., … Franco, D. W. (2004). Comparison between Cachaca and Rum Using Pattern

Recognition Methods. Journal of Agricultural and Food Chemistry, 52, 3429–3433.

https://doi.org/10.1021/jf035262+

Cutzach, I., Chatonnet, P., Henry, R., & Dubourdieu, D. (1997). Identification of Volatile

Compounds with a “Toasty” Aroma in Heated Oak Used in Barrelmaking. Journal of Agricultural

and Food Chemistry, 45, 2217–2224. https://doi.org/10.1021/jf960947d

da Silva, G. A., Maretto, D. A., Bolini, H. M. A., Teófilo, R. F., Augusto, F., & Poppi, R. J. (2012).

113

Correlation of quantitative sensorial descriptors and chromatographic signals of beer using

multivariate calibration strategies. Food Chemistry, 134, 1673–1681.

https://doi.org/10.1016/j.foodchem.2012.03.080



488. https://doi.org/10.1021/jf0511190

de Souza, P. P., Augusti, D. V, Catharino, R. R., Siebald, H. G. L., Eberlin, M. N., & Augusti, R.

(2007). Differentiation of Rum and Brazilian Artisan Cachaca via Electrostray Ionization Mass

Spectrometry Fingerprinting. Journal of Mass Spectrometry, 42, 1294–1299.

https://doi.org/10.1002/jms.1197

Dong, J. J., Li, Q. L., Yin, H., Zhong, C., Hao, J. G., Yang, P. F., … Jia, S. R. (2014). Predictive

analysis of beer quality by correlating sensory evaluation with higher alcohol and ester

production using multivariate statistics methods. Food Chemistry, 161, 376–382.


Fazzalari, F. A. (Ed.). (1978). Compilation of Odor and Taste Threshold Values Data. Baltimore, MD:

American Society for Testing and Materials.



Agricultural and Food Chemistry, 64, 637–645. https://doi.org/10.1021/acs.jafc.5b05426



Chemistry, 64, 9041–9053. https://doi.org/10.1021/acs.jafc.6b04046

Genthner, E. R. (2014). Identification of Rotundone as an Important Contributor to the Flavor of Oak Aged

Spirits. University of Illinois at Urbana Champaign. Retrieved from


González Álvarez, M., González-Barreiro, C., Cancho-Grande, B., & Simal-Gándara, J. (2011).

Relationships between Godello white wine sensory properties and its aromatic fingerprinting

114

obtained by GC–MS. Food Chemistry, 129, 890–898.


Granato, D., de Oliveira, C. C., Caruso, M. S. F., Nagato, L. A. F., & Alaburda, J. (2014). Feasibility

of different chemometric techniques to differentiate commercial Brazilian sugarcane spirits

based on chemical markers. Food Research International, 60, 212–217.

https://doi.org/10.1016/j.foodres.2013.09.044

Green, J. A., Parr, W. V., Breitmeyer, J., Valentin, D., & Sherlock, R. (2011). Sensory and chemical

characterisation of Sauvignon blanc wine: Influence of source of origin. Food Research

International, 44(9), 2788–2797. https://doi.org/10.1016/j.foodres.2011.06.005

Jung, H., Lee, S. J., Lim, J. H., Kim, B. K., & Park, K. J. (2014). Chemical and sensory profiles of

makgeolli, Korean commercial rice wine, from descriptive, chemical, and volatile compound

analyses. Food Chemistry, 152, 624–632. https://doi.org/10.1016/j.foodchem.2013.11.127

Kelley, L. E. (2014). Analysis of Potent Odorants in Spearmint Oils. University of Illinois at Urbana-

Champaign. Retrieved from https://www.ideals.illinois.edu/handle/2142/50415

Kotseridis, Y., Baumes, R., & Skouroumounis, G. K. (1998). Synthesis of labelled [ 2 H 4 ] β-

damascenone , [ 2 H 2 ] 2-methoxy-3-isobutylpyrazine, [ 2 H 3 ]α-ionone, and [ H ]β-ionone,

quantification in grapes , juices and wines. Journal of Chromatography A, 824, 71–78.

Kulkarni, P. P., Kadam, A. J., Mane, R. B., Desai, U. V., & Wadgoankar, P. P. (1999). Demethylation

of Methyl Aryl Ethers using Pyridine Hydrochloride in Solvent-free Conditions under

Microwave Irradiation. Journal of Chemical Research, S, 394–395. Retrieved from

http://pubs.rsc.org/EN/content/articlehtml/1999/jc/a901278h

Lahne, J. (2010). Aroma Characterization of American Rye Whiskey by Chemical and Sensory Assays.

University of Illinois at Urbana-Champaign. Retrieved from


Lapsongphon, N., Yongsawatdigul, J., & Cadwallader, K. R. (2015). Identification and

Characterization of the Aroma-Impact Components of Thai Fish Sauce. Journal of Agricultural

and Food Chemistry, 63, 2628–2638. https://doi.org/10.1021/jf5061248

115



Liu, J., Toldam-Andersen, T. B., Petersen, M. A., Zhang, S., Arneborg, N., & Bredie, W. L. P.

(2015). Instrumental and sensory characterisation of Solaris white wines in Denmark. Food

Chemistry, 166, 133–142. https://doi.org/10.1016/j.foodchem.2014.05.148

López, R., Aznar, M., Cacho, J., & Ferreira, V. (2002). Determination of minor and trace volatile

compounds in wine by solid-phase extraction and gas chromatography with mass spectrometric

detection. Journal of Chromatography A, 966, 167–177. https://doi.org/10.1016/S0021-

9673(02)00696-9

Lorjaroenphon, Y. (2012). Identification and Characterization of Potent Odorants Responsible for Typical and

Storage-Induced Flavors of Cola-Flavored Carbonated Beverages. University of Illinois at Urbana-

Champaign. Retrieved from https://www.ideals.illinois.edu/handle/2142/31932

Maga, J. A. (1989). The Contribution of Wood to the Flavor of Alcoholic Beverages. Food Reviews

International, 5(1), 39–99. https://doi.org/10.1080/87559128909540844

Mimura, N., Isogai, A., Iwashita, K., Bamba, T., & Fukusaki, E. (2014). Gas chromatography/mass

spectrometry based component profiling and quality prediction for Japanese sake. Journal of

Bioscience and Bioengineering, 118(4), 406–414. https://doi.org/10.1016/j.jbiosc.2014.04.006

Niu, Y., Zhang, X., Xiao, Z., Song, S., Eric, K., Jia, C., … Zhu, J. (2011). Characterization of odor-

active compounds of various cherry wines by gas chromatography-mass spectrometry, gas

chromatography-olfactometry and their correlation with sensory attributes. Journal of

Chromatography B: Analytical Technologies in the Biomedical and Life Sciences, 879(23), 2287–2293.

https://doi.org/10.1016/j.jchromb.2011.06.015

Otsuka, K., Zenibayashi, Y., Itoh, M., & Totsuka, A. (1974). Presence and Significance of Two

Diastereomers of β‐Methyl‐γ‐ octalactone in Aged Distilled Liquors. Agricultural and Biological

Chemistry, 38(3), 485.



116


Poisson, L., & Schieberle, P. (2008). Characterization of the Key Aroma Compounds in an

American Bourbon Whisky by Quantitative Measurements, Aroma Recombination, and

Omission Studies. Journal of Agricultural and Food Chemistry, 56(14), 5820–6.

https://doi.org/10.1021/jf800383v

Rayne, S., & Eggers, N. J. (2007). Quantitative determination of 4-ethylphenol and 4-ethyl-2-

methoxyphenol in wines by a stable isotope dilution assay. Journal of Chromatography A, 1167,

195–201. https://doi.org/10.1016/j.chroma.2007.08.049

Roberts, J. J., & Cozzolino, D. (2016). An Overview on the Application of Chemometrics in Food

Science and Technology—An Approach to Quantitative Data Analysis. Food Analytical Methods,

9(12), 3258–3267. https://doi.org/10.1007/s12161-016-0574-7

Robinson, A. L., Adams, D. O., Boss, P. K., Heymann, H., Solomon, P. S., & Trengove, R. D.

(2011). The relationship between sensory attributes and wine composition for Australian

Cabernet Sauvignon wines. Australian Journal of Grape and Wine Research, 17, 327–340.

https://doi.org/10.1111/j.1755-0238.2011.00155.x

Sampaio, O. M., Reche, R. V., & Franco, D. W. (2008). Chemical Profile of Rums as a Function of

their Origin. The Use of Chemometric Techniques for their Identificatio. Journal of Agricultural

and Food Chemistry, 56, 1661–8. https://doi.org/10.1021/jf0726841

Schneider, S., & Rolando, C. (1992). One Step Synthesis of Vanillin d3 (4-hydroxy-3-(methoxy d3)-

benzaldehyde). Journal of Labelled Compounds and Radiopharmaceuticals, 31(6), 489–492.

Schuh, C., & Schieberle, P. (2006). Characterization of the Key Aroma Compounds in the Beverage

Prepared from Darjeeling Black Tea : Quantitative Differences between Tea Leaves and

Infusion and Infusion. Journal of Agricultural and Food Chemistry, 54, 916–924.

Seisonen, S., Vene, K., & Koppel, K. (2016). The current practice in the application of

chemometrics for correlation of sensory and gas chromatographic data. Food Chemistry, 210,

530–540. https://doi.org/10.1016/j.foodchem.2016.04.134

117

Steinhaus, M., & Schieberle, P. (2005). Characterization of Odorants Causing an Atypical Aroma in

White Pepper Powder (Piper nigrum L.) Based on Quantitative Measurements and Orthonasal

Breakthrough Thresholds Pipernigrum. Journal of Agricultural and Food Chemistry, 53, 6049–6055.

https://doi.org/10.1021/jf0506030

Todeschini, R., Ballabio, D., Rojas, C., Manganaro, A., Mauri, A., Consonni, V., & Tripaldi, P.

(2007). Evaluation of Multivariate Classification and Variable Selection for the Distinction of

Cachaça and Rum Samples. Alimentos Ciencia E Ingeniería, 16(2), 113–115.

Uselmann, V., & Schieberle, P. (2015). Decoding the Combinatorial Aroma Code of a Commercial

Cognac by Application of the Sensomics Concept and First Insights into Differences from a

German Brandy. Journal of Agricultural and Food Chemistry, 63, 1948–1956.

https://doi.org/10.1021/jf506307x

Vilanova, M., Genisheva, Z., Masa, A., & Oliveira, J. M. (2010). Correlation between volatile

composition and sensory properties in Spanish Albariño wines. Microchemical Journal, 95, 240–

246. https://doi.org/10.1016/j.microc.2009.12.007

Willner, B., Granvogl, M., & Schieberle, P. (2013). Characterization of the Key Aroma Compounds

in Bartlett Pear Brandies by Means of the Sensomics Concept. Journal of Agricultural and Food

Chemistry, 61, 9563–9593.

Wiśniewska, P., Śliwińska, M., Dymerski, T., & Wardencki, N. (2017). Comparison of an Electronic

Nose Based on Ultrafast Gas Chromatography, Comprehensive Two-Dimensional Gas

Chromatography, and Sensory Evaluation for an Analysis of Type of Whisky. Journal of

Chemistry, 1–13. https://doi.org/https://doi.org/10.1155/2017/2710104 Research

Xiao, Z., Liu, S., Gu, Y., Xu, N., Shang, Y., & Zhu, J. (2014). Discrimination of Cherry Wines Based

on Their Sensory Properties and Aromatic Fingerprinting using HS-SPME-GC-MS and

Multivariate Analysis. Journal of Food Science, 79(3). https://doi.org/10.1111/1750-3841.12362

Reprinted with permission from John Wiley and Sons: Ickes, C. M., Lee, S.-Y., & Cadwallader, K. R.

(2017). Novel Creation of a Rum Flavor Lexicon Through the Use of Web-Based Material.

Journal of Food Science, 82(5), 1213–1223. https://doi.org/10.1111/1750-3841.13707

118

Chapter 5: Novel Creation of a Rum Flavor Lexicon Through the

Use of Web-Based Material

5.1 Abstract

Flavor lexicons help both manufacturers and consumers communicate the intricacies of flavor

nuances they experience within a product. Lexicon development typically requires the use of a

trained sensory panel to evaluate a representative sample set of the product category to generate

terms that describe certain product attributes. In the case of rum, there is considerable variation in

terms of style, flavor characteristics, and the sheer number of rums produced making it difficult to

create a lexicon in this manner. Furthermore, sensory fatigue from the high alcohol content can also

hinder lexicon development. This is the first study to create a rum flavor lexicon using web-based

material (comprising blogs, company descriptions and review websites) to minimize the time and

cost and to allow for the inclusion of a greater number of rum products. Reviews for over one

thousand different rums were utilized, comprising evaluations that described an array of rums,

including white, gold, aged and agricole. Each evaluation was coded for aroma, aroma-by-mouth,

and taste attributes using NVivoTM software to amass the sensory terms. Word frequency analysis

was conducted on coded attributes. The analysis yielded 147 terms, sorted into 22 different

categories. The most prominent terms included vanilla, oak, caramel, fruity, molasses and baking

spices.

Results of this study demonstrate that web-based material can be used for products containing a

large variation and where significant sensory fatigue is an issue to create a lexicon provided enough

evaluations of the product exist. The developed lexicon will aid in term generation for future

descriptive analysis panels on rum, as well as provide a standardized language for use in the rum

industry when training and evaluating samples for quality assurance as well as marketing.

119

5.2 Introduction

Flavor lexicons are standardized vocabularies that aid in the communication of the perceived

sensory attributes in foods and beverages. An established vocabulary provides terminology to

describe flavor perceptions in a consistent manner. Lexicons aid in communication between

researchers, product developers and manufacturers for use in quality assurance and product

assessment, as well as to aid marketers in articulating the subtle flavor perceptions to consumers.

Lexicons typically contain terminology for multiple sensory perceptions encompassed in the term

flavor, including aroma, aroma-by-mouth, taste, mouthfeel and trigeminal sensations.

Lexicons have been developed for a variety of different food products including spices (Lawless,

Hottenstein, & Ellingsworth, 2012), cheese (Drake, McInvale, Gerard, Cadwallader, & Civille, 2001),

bread (Kleinert, Bongartz, Raemy, & Wadenswil, 2009), olive oil (Mojet & de Jong, 1994), almonds

(Civille, Lapsley, Huang, Yada, & Seltsam, 2010), tea (Koch, Muller, Joubert, van der Rijst, & Næs,

2012) and orange juice (Pérez-Cacho, Galán-Soldevilla, Mahattanatawee, Elston, & Rouseff, 2008).

Additionally, a number of lexicons have been developed for alcoholic beverages including beer

(Clapperton, Dalgiesh, & Meilgaard, 1976; Meilgaard, Dalgliesh, & Clapperton, 1979; Parker, 2012),

brandy (Jolly & Hattingh, 2001), cognac (Lurton, Ferrari, & Snakkers, 2012), distilled beverages (Mc

Donnell, Hulin-Bertaud, Sheehan, & Delahunty, 2001), whisky (Lee, Paterson, Piggott, &

Richardson, 2001; Piggott & Jardine, 1979; Swan, Howie, & Burtles, 1981) and wine (Noble et al.,

1984, 1987). The wine flavor wheel is one of the most well-known lexicons, recognized by

connoisseurs and consumers alike. Standardized terms, definitions, and references for a product

allow users to describe and verbalize the differences among products within a category (Lawless &

Civille, 2013).

Typical development of a lexicon requires the use of a trained sensory panel. Panelists generally

undergo numerous hours of training on evaluating different food products through descriptive

analysis (Lawless & Civille, 2013) . To develop a complete lexicon, samples that encompass all of the

diverse aspects of a product category need to be evaluated. This can include not only products from

different manufacturers but also products produced in various geographical regions or having

different maturities to develop a comprehensive lexicon (Drake et al., 2001). Once the products to

120

be evaluated have been selected, panelists will generate terms to describe their perceptions,

determine appropriate chemical or food product references, and develop specific definitions for

each attribute. After the initial term generation, terms that are redundant are removed from the list.

Once the terms and references have been determined, the lexicon will be organized into a

comprehensive list. The developed lexicon can then be validated to demonstrate that the terms are

effective for distinguishing relevant product characteristics within the category (Lawless & Civille,

2013). Panelists start by scaling the attributes according to the selected references. Each sample is

then evaluated by the panelists with each attribute being rated in comparison to its corresponding

reference. The compiled data allow researchers to identify how products within the category differ

from each other.

Once a lexicon has been developed, it can be converted into a flavor wheel which provides a visual

representation of the generated terms (Lawless & Civille, 2013). Terms are typically grouped by

category, with the grouping identified in the inner circle and the specific term listed along the

outside. The completed wheel can be used to train new panelists by aiding with term selection and

communication with other scientists, consumers, or marketers.

A limited number of sensory studies have been conducted on rum. A vocabulary for evaluating

distilled spirits was developed using a white rum along with white tequila, vodka, gin, grappa, and

pear acquavite (Mc Donnell et al., 2001). Term development generated 100 initial terms which were

then reduced to 30 final terms for use in qualitative descriptive analysis. De Souza and others (de

Souza, Vásquez, del Mastro, Acree, & Lavin, 2006) used descriptive sensory analysis to compare the

aroma profiles of cachaça and rum on the basis of 10 aroma attributes. Franitza and others

(Franitza, Granvogl, & Schieberle, 2016a, 2016b) have recently published two papers where aroma

profile analysis was used to compare model aroma recombinates to the original rum samples. These

studies are limited in scope as they each evaluated only one or two rums, with a primary focus on

the flavor chemistry of the products rather than on their sensory profiles. The most comprehensive

sensory study on rum to date is a master’s thesis by Gómez (Gómez, 2002) where 33 terms were

generated through the evaluation of 15 different rums. However, this vocabulary is incomplete as

demonstrated by their later descriptive analysis panel which utilized terms that were not part of the

initial vocabulary (Gómez, 2002). The main emphasis of these studies has been to develop

terminology to aid in evaluating one or a specific subgroup of rum samples rather than to gain a

comprehensive understanding of rum flavor across the entire category.

121

Currently, there is no published flavor lexicon for rum. Rum is an extremely complex product

category due to its limited standards of identity. The only production requirement is that the

distillate must be produced from a sugarcane by-product (Labeling and advertising of distilled spirits,

27 C.F.R. § 5.22 1969). Since rum can be produced from a variety of sugarcane products, be distilled

in multiple ways, and aged in any type of barrel the manufacturer desires, there is a significant

amount of variation amongst products classified as rum. Rums are typically grouped into several

categories including white, gold, aged, agricole, flavored and spiced rums (Ayala, 2001). White rums

are typically aged in either stainless steel tanks or wood casks for 1 to 2 years and then filtered

through charcoal. Gold rums are typically aged for 1 to 3 three years in wood casks. Aged rums are

those that have been aged for at least 3 years or more. Agricole rums are produced from pure sugar

cane juice rather than molasses and are produced predominately in the French West Indies. Flavored

and spiced rums are rums blended with fruits and spices during production and should not be

compared with rums produced in the traditional manner. Due to the enormous variation in the

product category and the sensory fatigue panelists would experience from the high alcohol content

of the product, a rum lexicon developed in the traditional way would be both expensive and time

intensive.

Web-based materials are increasingly available, with more content being added to the internet on a

daily basis. The ease of access allows enthusiasts and connoisseurs of spirits and alcoholic beverages,

such as rum, to publish their opinions and reviews online. Additionally, companies readily market

their product through the use of descriptive terminology. Compiling the terms already being used by

consumers and the rum industry provides an excellent starting point for the creation of a flavor

lexicon. A similar approach has recently been used to create a flavor wheel for Chenin Blanc wine,

where the researchers used the terminology found in the John Platter wine guide to construct the

wheel (Valente, 2016).

With this in mind, the objective of this study was to evaluate a new method for lexicon creation of

rum through the use of web-based materials. Rum products have numerous product descriptions

and reviews available online. Use of web-based material allowed for the collection of data on a wide

variety of rum products, more than could feasibly be evaluated with traditional methods.

122


Selection of Web-based Material

Web-based material consisting of blogs, company descriptions and review websites, were chosen

based on the following criteria: use of qualitative flavor descriptors and website organization. Web-

material was found through the use of GoogleTM search using keywords such as “rum blogs” and

“rum reviews.” Blogs and websites that provided only reviews and/or ratings for rums but no

qualitative flavor descriptions were excluded from the study. Company web pages were sought for

each rum reviewed on other websites. All companies that had web pages and product descriptions

were included. Seventeen websites and 57 company descriptions were selected for analysis (Table

5.1). Only reviews available through June 2015 were included. All evaluations were captured as

PDFs and imported into NVivoTM (NVivo qualitative data analysis software; QSR International Pty

Ltd. Version 10, 2014).

Evaluation of Web-Based Material for Sensory Terms

Each rum evaluation was coded by hand for aroma (orthonasal perception), aroma-by-mouth

(retronasal perception), and taste (consisting of basic tastes, mouthfeel and trigeminal) attributes.

Only qualitative descriptors were coded. Subjective descriptors such as good, bad, excellent, and

mature were excluded. Once all evaluations were coded, word frequency analysis was conducted in

NVivo on the collected attributes. Any descriptor that appeared at least ten times throughout the

entire dataset was selected for the preliminary lexicon. This was a natural cutoff in the data as there

were still sensory relevant terms and each term encompassed 0.05% of the total terms coded.

Categorization of Flavor Terms

A sorting exercise was performed to categorize the terms. Ten individuals, 3 male, and 7 female, 23-

29 years of age were selected from University of Illinois students based on availability and interest in

participating on the sorting panel. Each person was given a bag containing all of the selected terms

written on paper slips. Participants were instructed to sort the terms into categories as they saw fit,

based solely on their knowledge of the presented terms. Panelists were allowed to ask for definitions

of terms they did not know. They were also instructed to group redundant terms and remove terms

they felt were coded by mistake or not relevant to flavor perceptions such as palate, little, slightly,

123

and sip. Once the participants had sorted all of the terms, they were asked to label and write a brief

description for each category they constructed. The data were visually analyzed by the panel

facilitator who compiled the categories developed from panelists into one comprehensive lexicon.

Terms that were grouped together consistently by multiple assessors were placed together in the

final lexicon. Terms that were sorted into multiple categories by panelists with no clear majority

were placed into their final grouping based on the organization of previous alcoholic beverage flavor

wheels (Noble and others 1984, 1987; Jolly and Hattingh 2001; Lee and others 2001).


Rum reviews from 17 different websites as well as product descriptions from 57 different companies

were compiled for data analysis. Over 3,000 individual reviews were coded for 1,053 different

products which encompassed white, gold, aged, and agricole rums. Flavored and spiced rums were

not included in this analysis as additional flavoring agents are added to these products to change the

flavor of the rum. Initially, 267 individual terms were selected through the word frequency analysis.

The ten most frequently used terms to describe rum flavor were vanilla, sweet, spices, oak, caramel,

fruity, dry, smooth, sugar, and molasses. These ten words comprised 28.48% of the total descriptors

coded.

The selected terms were given to participants for categorization. Through this exercise, a number of

terms were eliminated from the lexicon as they were either duplicate words that had multiple

spellings or endings such as fruit, fruits, and fruity or words that were not useful attribute

descriptors such as deep and tones. Terms which could be considered compound terms, or terms

composed of multiple attributes, were not eliminated from the lexicon. The individual

categorizations were then combined based on how often terms were grouped together. After term

categorization, a final lexicon consisting of 147 terms sorted into 22 categories was created (Table 2).

Descriptors were not identified as positive or negative with regard to overall flavor quality. All terms

remained as they appeared in the web-based material with the exception of brine which was replaced

with salty for the lexicon. The final lexicon is a compilation of the most prevalent terms currently

used to describe rum flavor. It is expected that only a subset of these terms would be required to

describe an individual rum.

124

Rum Flavor Wheel

The terms selected for the lexicon were organized into a flavor wheel for ease of use (Figure 1). The

rum flavor wheel is split into two tiers, with 22 first tier descriptors and 125 second-tier descriptors.

The innermost tier consists of the broader terms, while the outer wheel contains more precise

terminology within each inner category. Terms that are more similar are located closer to each other

on the wheel.

All of the first tier descriptors have been used in other alcoholic beverage lexicons as category

groupings except for chocolate, coffee and confections (Clapperton and others 1976; Meilgaard and

others 1979; Shortreed 1979; Noble and others 1984, 1987; Langstaff and Lewis 1993; Jolly and

Hattingh 2001; Lee and others 2001; Le Barbe 2003; Lurton and others 2012). Terms associated

with sugary products are typically categorized as sweet-associated rather than categorized under

sugar as we have done (Shortreed 1979; Jolly and Hattingh 2001; Lee and others 2001; Le Barbe

2003). All of the categories have been used in at least two other lexicons except alcoholic beverages

and dairy products. The whisky flavor wheel published by Lee and others (2001) is the only lexicon

to include other alcoholic beverages as a category which they label “previous use.” Additionally,

mouthfeel and trigeminal sensations tend to be grouped together when found on previous flavor

wheels (Clapperton and others 1976; Meilgaard and others 1979; Shortreed 1979; Noble and others

1984, 1987; Langstaff and Lewis 1993; Jolly and Hattingh 2001; Lee and others 2001; Le Barbe 2003;

Lurton and others 2012). Meanwhile, chocolate and coffee terms have not previously been used as a

category, although they are found as second tier terms (Clapperton and others 1976; Noble and

others 1984, 1987; Jolly & Hattingh 2001; Le Barbe 2003; Lurton and others 2012). Chocolate has

previously been grouped under caramel, smoky, sweet associated and vanilla, while coffee has been

grouped under caramel, smoky, toasted, and woody.

Terms Most Common to Rum Categories

The abundance of data collected using web-based material allows for the identification of not only

the terms used to describe rum as an entire class but also demonstrates how the various categories

of rum differ from each other. Rums categorized as white, gold, and a variety of age statement types

including 2-5 years, 5-10 years, 10-20 years, and 20 plus years were able to be analyzed. The 15 most

prevalent aroma and aroma-by-mouth terms in each categorization can be found in Table 3. These

125

data indicate that all rums appear to be characterized by vanilla, caramel, fruity and spices. White

rums seem to be further characterized by coconut, floral, banana and mineral notes compared to the

other categories; whereas, woody, tobacco, toffee and chocolate notes seem to increase as a function

of age. Previous sensory profiles of rums also show the terms vanilla, caramel, fruity and terms

related to spices as key attributes to describe rum flavor along with alcoholic/ethanolic and woody

(Gómez 2002; De Souza and others 2006; Franitza and others 2016a, b). Previous studies have not

focused on aroma differences as a result of maturation, making this the first study to provide insight

into how rum aroma profiles may change as a result of maturation.

Discussion of the Rum Lexicon

The use of web-based material for term generation provides an excellent starting point for lexicon

development. The final lexicon contains many terms that overlap with already published lexicons for

alcoholic beverages. All of the category descriptors except trigeminal, chocolate, leather, dairy and

coffee as well as more than half the individual terms have been used in previous lexicons for

alcoholic beverages (Piggott and Jardine 1979; Swan and others 1981; Noble and others 1984; Jolly

and Hattingh 2001; Lee and others 2001; Mc Donnell and others 2001; Schmelzle 2009). The

substantial overlap of terminology indicates that appropriate descriptors were coded for and used to

develop the wheel. The terms not previously used in flavor lexicons are likely to be descriptors that

are unique to rum.

A similar data mining approach has recently been used to construct a flavor wheel for Chenin Blanc

wines (Valente 2016). The researchers collected terms from the John Platter Wine Guide over a

seven year period. The guide annually rates the wines produced by the vineyards in South Africa,

enabling them to amass terms from 2746 wines. Our approach is similar in that both studies are

collecting potential terms from previously published reviews. Our study further expands the use of

data mining, as we have evaluated reviews from multiple online sources rather than just a single

publication. Both studies demonstrate that preexisting data can be used to construct a valid flavor

lexicon.

While the current practice for lexicon development makes use of a descriptive analysis panel,

original flavor wheels for beverages such as wine and beer were developed through the use of

experts in the field who suggested terms to be included in the lexicon (Clapperton and others 1976;

126

Noble and others 1987). This collaboration led to the creation of lexicons which adequately

described the beverage class. A similar approach was followed here, where reviews from rum

experts, enthusiasts, and connoisseurs were collected and evaluated to create the lexicon. While the

reviewers did not specifically propose terms, we used their descriptors in the development of the

rum lexicon.

A limiting factor of the lexicon is that web-reviewers do not adequately define the terms they use in

their descriptions. As a result, no precise definitions or references for terms were collected to create

this lexicon. While definitions are helpful for a better grasp of the accurate perception of an

attribute, many flavor wheels do not contain definitions for their terms. This is true for many whisky

wheels (Jackson 1989; MacLean 1997) as well as the McCormick Spice Wheel (Lawless and others

2012). While definitions are helpful, they are not essential for the wheel to be useful. Many times

panelists identify that a sample contains a familiar attribute but are not able to articulate what they

perceive without the visual cue to accompany the perception. Having a lexicon accessible to them

gives panelists a selection of words to choose from, instead of forcing them to generate the term

from memory. Additionally, a majority of the terms included in the rum lexicon have inherent

meanings for panelists such as the aroma of caramel, clove or banana.

While definitions are not always a necessity for connoisseurs or expert judges, they are essential if

the terms are to be used as part of a descriptive analysis panel. Definitions, as well as specific

references, will need to be developed and selected in order for a panel to consistently rate the terms

across samples. Defining terms through the use of a descriptive analysis panel would be beneficial to

clarify the meaning of the terms with respect to how they are perceived in rum. This would be

especially useful for the trigeminal and mouthfeel terms, as people tend to be less familiar with those

sensations compared to terms which describe food products.

Another limiting factor of the rum lexicon is the use of compound terms. During typical lexicon

development, panelists are discouraged from using terms that combine multiple attributes and

prompted to break those terms apart into singular attributes. However, this is not the case for web

reviewers, they describe what they perceive, and sometimes the attributes are combined to form a

complex term that is more easily identifiable than the individual attributes. For example, our wheel

contains the term mocha which combines the aromas of chocolate and coffee, and even those

attributes could be considered compound terms as well. Praline is another compound term of nutty

127

and caramel. Additional compound terms encompassed in the rum lexicon include marzipan, crème

brûlée, tea, cigar, fruitcake, and gingerbread. Regardless, compound terms have been used in other

lexicons as well. Most notably, coffee and tea are typically included in lexicons (Noble and others

1984, 1987; Jolly and Hattingh 2001; Lee and others 2001; Le Barbe 2003; Lurton and others 2012)

even though entire lexicons have been created to characterize those products (Seo and others 2009;

Koch and others 2012; Muller and Joubert 2013; Theron and others 2014; Spencer and others 2016).

These terms are most likely included in lexicons as they are easy for evaluators to identify since the

combination of aromas has a characteristic profile that people encounter on a regular basis. The

terms marzipan and cigar (box) have been included in the whisky (Lee and others 2001) and cognac

(Lurton and others 2012) flavor wheels, respectively. Additionally, the whisky (Lee and others 2001),

sparkling wine (Le Barbe 2003), and brandy (Jolly and Hattingh 2001) flavor lexicons include

alcoholic beverages such as sherry and port as attributes.

It should be noted that to the best of our knowledge this is the first attempt to create a flavor

lexicon for rum. Even though over 3,000 reviews were utilized for this analysis, it is expected that

the lexicon will need to be modified over time, similar to the case for the wine and beer lexicons and

most recently the coffee lexicon (Meilgaard and other 1979; Noble and others 1987; Spencer and

others 2016). The beer terminology wheel was altered after discussions at the annual meetings of the

European and American brewing societies (Meilgaard and others 1979). Participants were able to

submit suggestions by mail in order to refine the lexicon to better meet the practical needs of

brewers’ societies (Meilgaard and others 1979). The wine aroma wheel also allowed for feedback

from experts in the field leading to the addition of several terms, reorganization of the wheel, and

the addition of chemical or food references for each term (Noble and others 1987). The new coffee

wheel utilized industry experts to develop the lexicon terms, followed by statistical analysis of term

sorting (Spencer and others 2016). These modifications have allowed the lexicons to change to be

better aligned with the needs of these ever-changing industries.

Conclusion

While the rum wheel includes terms utilized by companies and avid spirit enthusiasts, it is likely that

there are missing terms that may be identified by a trained panel. Validation of the wheel using a

descriptive analysis panel would be a first step in the refinement process. Additionally, this wheel

128

only includes terms which describe the finished rum products. The wheel would likely need to be

expanded if it were to be used during the production process.

The developed rum flavor lexicon provides manufacturers and retailers with an initial vocabulary to

describe rums. This will allow the same perceptions to be communicated consistently and provide

terminology for attributes that people may not have been able to previously articulate. The next step

would be to define the different terms using a descriptive analysis panel so that each term has a

standard reference and definition that can then be used to train new people for quality assurance.

This rum lexicon will be used to aid in the development of terms for a descriptive analysis panel by

providing a bank of terms for the panel to assess when generating terms.

129

5.5 Tables and Figure

Table 5.1 Web-based material used to create the rum flavor lexicon

Name of Blog Type Website Address Best of Luxury Blog http://www.bestofluxury.com/rankings-of-best-rum-liquors Bilgemunky Blog http://www.bilgemunky.com/category/pirate-reviews/rum/ Dowd's Tasting Notes Blog http://dowdtastingnotes.blogspot.com/ El Machete's Rum Reviews Blog http://macheterum.blogspot.com/ Fahrenheit 173 Blog http://fahrenheit173.com/ratings/rum/?search=&top=all&origin=all&type=w Got Rum Blog http://www.gotrum.com/rumreviews Proof66 Blog http://www.proof66.com/liquor/rum.html Refined Vices Blog Blog https://refinedvices.com/rum-reviews/ Rob's Rum Guide Blog http://www.robsrum.com/ Rum Dood Blog http://rumdood.com/rum-reviews/ Rum Ratings Blog https://www.rumratings.com/brands Scottes' Rum Pages Blog https://scottesrum.com/category/all-rum-reviews/ Spirits Review Blog http://spiritsreview.com/class/rum/ Tastings Blog http://www.tastings.com/Home.aspx The Rum Howler Blog https://therumhowlerblog.com/rum-reviews/ The Rum Shop Blog http://www.rumshop.net/ The Rumelier Blog http://www.therumelier.com/index.html Admiral Rodney Company http://www.saintluciarums.com/admiral-rodney.html A.H. Riise Company https://www.ahriiserum.com/products.html Amrut Company http://www.amrutdistilleries.com/validated/pages/opdr.html Angostura Company http://www.angostura.com/Brands/ Antigua Distillery Cavalier Company http://www.antiguadistillery.com/our-range.html Appleton Estate Company http://www.appletonestate.com/en/our-rum/ Atlantico Company http://www.atlanticorum.com/services3 Bacardi Company https://www3.bacardi.com/us/en/our-rums/ Ron Barcelo Company http://www.ronbarcelo.com/initial.html Brugal Company https://www.brugal-rum.com/our-rum/ Cana Brava Company http://www.canabravarum.com/ Clarke's Court Company http://www.clarkescourtrum.com/products/aged Rhum Clement Company http://rhumclementusa.com/index.php Chairman's Company http://www.saintluciarums.com/chairmans-reserve.html Cockspur Company http://www.cockspurrum.com/concocktions/ Cruzan Company http://www.cruzanrum.com/rum-collection/ Deadhead Company http://deadheadrum.com/the-rum.html Dictador Company http://www.dictador.com/dictador-rums.html Diplomatico Company http://www.rondiplomatico.com/rums Don Pancho Origenes Company http://origenesdonpancho.com/?age-verified=cd956f803a Don Papa Company http://www.donpaparum.com/the-goods.html Dos Maderas Company http://www.rondosmaderas.com/en/index.php El Dorado Company http://theeldoradorum.com/our-portfolio Facundo Company http://www.facundorum.com/the-rum-collection/ Flor de Cana Company http://flordecana.com/eng/products Gosling's Company http://www.goslingsrum.com/our-products/ Green Island Company http://greenislandrum.com/products.html Havana Club Company http://havana-club.com/en/cuban-rum/havana-club-rum Kirk and Sweeney Company http://www.3badge.com/kirkandsweeney/ Koloa Company http://koloarum.com/rums/ Montanya Company http://www.montanyarum.com/home-rums/ Mount Gay Company http://www.mountgayrum.com/ Neisson Company http://www.neisson.fr/-rubrique14-.html?lang=en New Grove Company http://www.newgrove.mu/# Newfoundland Screech Company http://screechrum.com/#products Opthimus Company http://spiritimportsinc.com/brands/opthimus/ Penny Blue Company http://indianoceanrum.com/pennybluerum/ Plantation Company http://www.plantationrum.com/rums-plantation#vintages Rhum Agricole Bologne Company http://www.rhumbologne.fr/en/rums.html Rhum Barbancourt Company barbancourtrhum.com/the-rhums/

130

Table 5.1 (cont.) Web-based material used to create the rum flavor lexicon

Name of Blog Type Website Address Rhum J.M Company http://www.rhumjmusa.com/index.php Ron Abuelo Company http://www.ronabuelopanama.com/category/brand/?lang=en&lang=en Ron Matusalem Company http://www.matusalem.com/en.html Rum Nation Company http://www.rumnation.com/en/ Saint James Company http://www.saintjames-rum.com/#!/gamme Santa Teresa Company https://ronsantateresa.com/us/#/we-make-rum Sergeant Classick Company http://sgtclassick.com/default.asp St. Barth Company http://www.rhumstbarth.com/collection-rhum.asp Tanduay Company http://tanduay.com/brands/tanduay-rhum-5-years/ The Real McCoy Company http://www.realmccoyspirits.com/therum TOZ Company http://www.saintluciarums.com/toz.html Trois Rivieres Company http://www.plantationtroisrivieres.com/3-rivieres/the-collection/id/1319 Wild Geese Company http://thewildgeesecollection.com/rum/our-rum/ Yacuro Company http://www.calderonglobaltrade.com/en/ron_yacuro/

131

Table 5.2 Categorization of terms generated from web-material used to create the flavor wheel

First Tier Second Tier First Tier Second Tier First Tier Second Tier

Sugar Candied Woody Barrel Citrus Lemon Honey Cedar Lime Honeycomb Oak Marmalade Maple Sandalwood Orange Marshmallow Marzipan Nutty Almond Alcoholic Arrack Molasses Hazelnut Bourbon Nougat Pecan Brandy Sugarcane Walnut Cognac Syrup Sherry Treacle Leather Saddle Whiskey Wine Caramel Brown Earthy Cigar Butterscotch Musty Medicinal Alkali Caramelized Peat Copper Cola Root Metallic Crème Brulee Sap Mineral Praline Tobacco Rubber Toffee Tar Vegetal Floral Dairy Buttery Fresh Tastes Bitter Cream Grass Salty Custard Green Sweet Herbal Chocolate Cocoa Minty Mouthfeel Astringent Fudge Tea Buttery Creamy Coffee Mocha Fruity Apple Crisp Apricot Dry Confections Fruitcake Cherry Heavy Gingerbread Currants Oily Meringue Dates Rough Figs Sharpness Baking Spices Allspice Grapes Silky Cardamom Peaches Smooth Cinnamon Pear Supple Clove Syrupy Ginger Dried Fruit Dried Apricot Thick Licorice Dried Dates Velvety Mace Dried Currants Nutmeg Dried Figs Trigeminal Bite Peppery Prunes Burn Vanilla Raisin Harsh Hot Roasted Burnt Tropical Fruit Banana Pungent Charred Coconut Sharp Smoky Mango Warm Toasted Pineapple Tannins

132

Table 5.3 Top 15 descriptors for each category of rum

White Gold 2-5 years 5-10 years 10-20 years 20+ years

Vanilla Vanilla Spices Vanilla Vanilla Vanilla

Banana Spices Vanilla Oak Oak Spices

Citrus Sugar Caramel Caramel Spices Oak

Caramel Caramel Oak Spices Caramel Fruity

Fruity Oak Fruity Fruity Fruity Caramel

Coconut Fruity Molasses Sugar Molasses Orange

Oak Molasses Sugar Molasses Toffee Sugar

Spices Banana Tobacco Cinnamon Sugar Toffee

Molasses Toffee Orange Tobacco Orange Woody

Sugar Woody Toffee Toffee Tobacco Molasses

Butterscotch Nutty Banana Orange Cinnamon Cinnamon

Cream Tobacco Coconut Butterscotch Woody Tobacco

Floral Orange Cinnamon Leather Banana Chocolate

Mineral Honey Butterscotch Banana Chocolate Honey

133

Figure 5.1 Rum flavor wheel developed from web-based material describing white, gold and aged rums

134

5.6 References

Ayala, L. K. (2001). The Rum Experience. Texas, USA: Rum Runner Press Inc.

Civille, G. V., Lapsley, K., Huang, G., Yada, S., & Seltsam, J. (2010). Development of an Almond

Lexicon To Assess the Sensory Properties of Almond Varieties. Journal of Sensory Studies, 25,

146–162. https://doi.org/10.1111/j.1745-459X.2009.00261.x

Clapperton, J. F., Dalgiesh, C. E., & Meilgaard, M. C. (1976). Progress Towards an International

System of Beer Flavour Terminology. Journal of the Institute of Brewing, 82, 7–13.



488. https://doi.org/10.1021/jf0511190

Drake, M. a, McInvale, S. C., Gerard, P. D., Cadwallader, K. R., & Civille, G. V. (2001).

Development of a Descriptive Language for Cheddar Cheese. Journal of Food Science, 66(9),

1422–1427.









Jolly, N. P., & Hattingh, S. (2001). A Brandy Aroma Wheel for South African Brandy. South African

Journal of Enology and Viticulture, 22(1), 16–21.

Kleinert, M., Bongartz, A., Raemy, E., & Wadenswil. (2009). Development of a Flavour Wheel for

Breads. Getreide, Mehl Und Brot, 63(1), 60–69.

Koch, I. S., Muller, M., Joubert, E., van der Rijst, M., & Næs, T. (2012). Sensory characterization of

135

rooibos tea and the development of a rooibos sensory wheel and lexicon. Food Research

International, 46, 217–228. https://doi.org/10.1016/j.foodres.2011.11.028

Labeling and advertising of distilled spirits, Title 27 Code of Federal Regulations, Pt. 5 Section 5:22.

1969 ed.

Lawless, L. J. R., & Civille, G. V. (2013). Developing Lexicons: A Review. Journal of Sensory Studies,

28(4), 270–281. https://doi.org/10.1111/joss.12050

Lawless, L. J. R., Hottenstein, A., & Ellingsworth, J. (2012). The McCormick Spice Wheel: A

Systematic and Visual Approach to Sensory Lexicon Development. Journal of Sensory Studies, 27,

37–47. https://doi.org/10.1111/j.1745-459X.2011.00365.x

Lee, K.-Y. M., Paterson, A., Piggott, J. R., & Richardson, G. D. (2001). Origins of Flavour in

Whiskies and a Revised Flavour Wheel: a Review. Journal of the Institute of Brewing, 107(5), 287–

313. https://doi.org/10.1002/j.2050-0416.2001.tb00099.x

Lurton, L., Ferrari, G., & Snakkers, G. (2012). Cognac: production and aromatic characteristics. In J.

Piggott (Ed.), Alcoholic Beverages: Sensory Evaluation and Consumer Research (pp. 242–266). Oxford:

Woodhead Publishing Limited. https://doi.org/10.1016/B978-0-85709-051-5.50011-0

Mc Donnell, E., Hulin-Bertaud, S., Sheehan, E. M., & Delahunty, C. M. (2001). Development and

Learning Process of a Sensory Vocabulary for the Odor Evaluation of Selected Distilled

Beverages Using Descriptive Analysis. Journal of Sensory Studies, 16(2001), 425–445.

Meilgaard, M. C., Dalgliesh, C. E., & Clapperton, J. F. (1979). Beer Flavour Terminology. Journal of

the Institute of Brewing, 85, 38–42.

Mojet, J., & de Jong, S. (1994). The sensory wheel of virgin olive oil. Grasas Y Aceites, 45(1–2), 42–

47.

Noble, A. C., Arnold, R. A., Buechsenstein, J., Leach, E. J., Schmidt, J. O., & Stern, P. M. (1987).

Modification of a Standardized System of Wine Aroma Terminology. American Journal of Enology

and Viticulture, 38(2), 143–146.

Noble, A. C., Arnold, R. A., Masuda, B. M., Pecore, S. D., Schmidt, J. O. S., & Stern, P. M. (1984).

Progress Towards a Standardized System of Wine Aroma Terminology. American Journal of

136

Enology and Viticulture, 35(2), 107–109. https://doi.org/10.1073/pnas.0703993104

Parker, D. K. (2012). Beer: production, sensory characteristics and sensory analysis. In J. Piggott

(Ed.), Alcoholic Beverages: Sensory Evaluation and Consumer Research (pp. 133–158). Oxford:

Woodhead Publishing Limited. https://doi.org/10.1016/B978-0-85709-051-5.50006-7

Pérez-Cacho, P. R., Galán-Soldevilla, H., Mahattanatawee, K., Elston, A., & Rouseff, R. L. (2008).

Sensory Lexicon for Fresh Squeezed and Processed Orange Juices. Food Science and Technology

International, 14(Suppl. 5), 131–141. https://doi.org/10.1177/1082013208094723

Piggott, J. R., & Jardine, S. P. (1979). Descriptive Sensory Analysis of Whisky Flavour. Journal of the

Institute of Brewing, 85(2), 82–85. https://doi.org/10.1002/j.2050-0416.1979.tb06830.x

Swan, J. S., Howie, D., & Burtles, S. M. (1981). Sensory and Instrumental Studies of Scotch Whisky

Flavour. The Quality of Foods and Beverages, 201–223. https://doi.org/10.1016/B978-0-12-

169101-1.50020-3

Valente, C. C. (2016). Understanding South African Chenin Blanc wine by using data mining techniques applied

to published sensory data. Stellenbosch University. Retrieved from

http://scholar.sun.ac.za/handle/10019.1/98866

137

Chapter 6: Characterization of Sensory Differences in Premium

Rums Through the Use of Descriptive Analysis Panel

6.1 Abstract

Rum is a highly diverse distilled spirit due to the numerous variations in methods used in its

products resulting from its limited standard of identify. This large product variation makes it difficult

to define typical rum flavor. Overall, sensory studies evaluating rum flavor are limited. The little

sensory work that has been done typically only focuses on one or two rum samples and only aroma

perception of rums has been evaluated. A comprehensive sensory analysis of premium rums is

lacking.

The aims of this study were to 1) identify and quantify the sensory differences between seven

premium rums and two mixing rums and 2) validate the developed rum flavor lexicon (Chapter 5).

Results showed that the use of the rum flavor lexicon aided in sensory term generation, as 17

additional terms were generated after the wheel was provided to panelists. Thirty-eight sensory terms

were generated and evaluated by the panel. Only five of the finalized terms were not found on the

flavor wheel. Twenty attributes were found to be significantly different between rums. DX and

DR12 were significantly different from the other seven rums, with higher perceptions of brown

sugar, caramel, vanilla, and chocolate aroma, caramel, maple, and vanilla aroma-by-mouth and

caramel aftertaste. The other seven rums had similar aroma profiles, with the mixing rums having

the lowest intensities for brown sugar, caramel, vanilla attributes.

6.2 Introduction

Rum is a complex and ill-defined spirit. With its only standard of identity being that it is produced

from a sugarcane by-product, numerous production variations have been developed. These

differences in production methods include variations in starting material, the length and type of

fermentation, distillation apparatus, type of barrels used for aging and length of maturation. The

138

resulting rum products are highly varied, making it difficult to identify and describe typical rum

flavor.

Sensory descriptive analysis (DA) methods enable the identification and quantification of sensory

differences between food samples (Lawless & Heymann, 1999b; Powers, 1988; Stone & Sidel, 1993;

Zook & Pearce, 1988). A number of different DA methods exist including Flavor Profile

(Cairncross & Sjöström, 1950; Caul, 1957), Quantitative Descriptive Analysis® (QDA) (Stone, 1992;

Stone, Sidel, Oliver, Woolsey, & Singleton, 1974), Spectrum TM Descriptive Analysis (Muñoz &

Civille, 1992), and Free-Choice Profiling (Williams & Langron, 1984). The QDA and spectrum

methods are the two most frequently used for descriptive analysis of products. QDA requires the

use of a trained sensory panel to evaluate the products (Stone & Sidel, 1993; Stone et al., 1974; Zook

& Pearce, 1988). Panelists are trained over multiple sessions, usually spanning several weeks. Panel

training consists of term generation, identification of appropriate references for the generated terms,

development of a specific definition for each term, scaling of the references, and practice rating the

samples using the developed references and reference scores (Stone & Sidel, 1993). During final

sample evaluations, panelists rate the samples multiple times in individual in booths using a line scale

with verbal anchors. Finally, statistical analysis is performed on the collected data set. The Spectrum

method is similar to QDA although panelists undergo more intense training and use a

predetermined list of sensory attributes and references to describe the products (Lawless &

Heymann, 1999b; Muñoz & Civille, 1992). Another key difference is that panelists use a numerical

15 point scale rather than a line scale. Both methods use panelists as trained instruments to quantify

the sensory of differences between the products.

Research characterizing the sensory properties of rum is limited. Previous studies have only

evaluated one modality (aroma) or only characterized one or two rum samples (de Souza, Vásquez,

del Mastro, Acree, & Lavin, 2006; Franitza, Granvogl, & Schieberle, 2016a, 2016b; Gómez, 2002;

Magnani, 2009). De Souza and others (de Souza et al., 2006) were the first to evaluate rum flavor.

They performed a limited descriptive analysis panel where 10 terms were generated to describe the

aroma differences between rum and cachaça. Results indicated that rum is higher in caramel and

vanilla aroma attributes while cachaça was higher in grassy, sulfury, vinegar, spicy, citrus, alcohol and

melon attributes.

139

Aroma profile analysis has also been used to characterize several rum samples. The first study

evaluated the aroma of two rum samples, a 15 year old solera system rum and a cheap (mixing) rum

(Franitza et al., 2016a). Only six attributes were chosen to describe the samples, ethanolic, malty,

butter-like, fruity, clove-like, and vanilla-like. Franitza then went on to characterize the flavor

changes that occur during the rum making process (Franitza et al., 2016b). Four different steps in

the production process were analyzed: the starting material (molasses), the mash, distillate and final

rum. Eight aroma attributes were evaluated, adding caramel-like and baked apple-like to their

previous descriptors.

Only one study has evaluated all sensory modalities (aroma, aroma-by-mouth, taste, mouthfeel, and

aftertaste) for rum (Magnani, 2009). Magnani used QDA to characterize the sensory difference

between two rum and two cachaça samples using 17 sensory attributes. The results indicated that

cachaça was higher in woody and sweet aroma and aroma-by-mouth as well as higher viscosity, body

and golden color.

While these studies give a basic insight into the defining sensory characteristics of rum, they give

little insight into the variations that exist in flavor between rum samples across the larger category.

These sensory studies focused on one or two samples or followed the production process of a single

type of rum. Due to the huge amount of variation amongst rums, sensory evaluation of a greater

number of rums are needed to gain a comprehensive understanding of rum as an entire class.

The most complete sensory study to date was a descriptive analysis panel done by Gomez (Gómez,

2002). Nine commercially available rums and one experimental sample were evaluated for 22 sensory

attributes and 4 chemical aroma sensations. Rums samples were chosen to include a variety of raw

materials, processing protocols, production regions and price points. All rum samples were able to

be distinguished from each other with samples having significant differences in woody, ethanol,

caramel, honey, smoke, vanilla, banana, prune, cardboard and ocean-like aroma attributes. This study

was the first to demonstrate the significant variation in rum aroma across a larger and more

complete rum sample set.

The previous work has begun to identify the characterizing sensory attributes of rum flavor as well

as differences that exist within the broader rum category. However, the large variation of products in

140

the market requires more sensory analysis of rum samples to gain a better understanding of overall

rum flavor.

The aim of this study was to evaluate and quantify the sensory differences of a variety of premium

aged rums and mixing rums as well as to validate the developed rum flavor lexicon (Chapter 5)

(Ickes, Lee, & Cadwallader, 2017). It was hypothesized that mixing rums would be easily

distinguishable from the premium rums. Additionally, that many of the terms used to describe the

difference in rum flavor would be found on the rum flavor wheel.


Materials

Nine commercially available rums, previously analyzed in chapters 3 & 4, were purchased at the local

liquor store (Champaign, IL). All nine rums, (Bacardi Superior [BW], Bacardi Gold [BG], Appleton

Estate V/X [AE], Appleton Estate Extra [AE12], Ron Abuelo: Añejo 7 years [RA7], Diplomatica

Reserva Exclusiva [DR12], El Dorado 12 year old [ED12], Ron Zacapa (Centenario) XO: Solera

Gran Reserva Especial [RZ], Dictador XO Insolent [DX]) had reported ethanol concentration of

40% alcohol by volume (ABV). Mention of the brand name of these rums does not imply any

research contact or sponsorship and is not for advertisement or endorsement purposes. Before

testing, 20 mL of each sample was measured into a black double old-fashion glass (Threshold,

Target, USA) and covered with a glass petri dish no more than one hour before the panel

assessment.

Attribute references were prepared daily and placed into 2 or 4 oz. lidded plastic soufflé cups (Dart

Container Corporation, Mason, MI) and labeled with the reference identity. References were not

prepared more than 24 h prior to evaluation. A complete list of attributes, definitions, references,

reference scores and preparation procedures can be found in Table 6.1. Product information for

references can be found in Appendix I.

The following materials were purchased for initial panelist taste screening: caffeine (Fisher Scientific,

Fair Lawn, NJ), citric acid (Ball, Hearthmark, LLC dba Jarden Home Brands, Daleville, IN), sodium

141

chloride (Morton Salt, Morton Salt, Inc., Chicago, IL), and sucrose (C&H Pure Cane Sugar, Domino

Foods, Inc., Yonkers, NY). Additionally, the following compounds were used for aroma screening:

ethyl butyrate, eugenol, oak lactone, 2-phenethyl alcohol, and vanillin. All compounds were food

grade and purchased from Sigma-Aldrich (St. Louis, MO).

Subject Recruitment

All materials related to panelist recruitment and compensation were approved by the Institutional

Review Board (IRB) at the University of Illinois Urbana-Champaign (IRB Protocol Number: 16854),

Appendix D. Panelists were recruited through the departmental listserve and by word of mouth.

Potential panelists completed an initial recruitment questionnaire to screen for availability and

familiarity with distilled beverages (Appendix E). After passing the initial questionnaire, potential

panelists then completed a sensory screening for basic taste perception and aroma identification.

Potential panelists were required to identify several basic taste solutions: caffeine (0.024% wt/vol),

citric acid (0.05% wt/vol), sodium chloride (0.1% wt/vol), and sucrose (0.7% wt/vol). The solutions

(10mL) were placed into 1 oz. lidded plastic cups. Panelists were also screened for their ability to

detect and correctly identify odorants common to rum: ethyl butyrate (1.03 mg/mL), eugenol (6.00

mg/mL), oak lactone (0.0396 mg/mL), 2-phenethyl alcohol (1.05 mg/mL), and vanillin (2.46

mg/mL). The odorants were spiked (10μL, 2 μL, 10 μL, 10 μL, and 5 μL respectively) onto an

odorless cotton ball and placed into a 5.5oz lidded plastic cup. All samples were labeled with random

three digit codes. The ballot used for screening can be found in Appendix F. Panelists were also

required to present a valid form of identification at the screening to verify that they were over 21

years of age. Panelists were selected if they achieved at least 70% acuity on the aroma identification

and were available at the time selected for the panel. Panelists granted informed consent (Appendix

G) before the start of the study.

Descriptive Analysis Procedure

The panel methodology was a hybrid of the Quantitative Descriptive Analysis® (Stone, 1992) and

the SpectrumTM method (Meilgaard, Civille, & Carr, 2007; Muñoz & Civille, 1992). The nine samples

were evaluated by eight panelists, 4 male and 4 female, ages 23-66, over a total of 23 days. On the

first day of panel, panelists were introduced to the specific descriptive analysis (DA) method used in

this study. The first six sessions (1 h each) consisted of term and reference generation, followed by

142

reference refinement. The panelists were presented with three random rum samples each day and

asked to generate terms for the attributes they perceived in the samples for aroma, aroma-by-mouth,

mouthfeel, taste and aftertaste modalities. Through group discussion, panelists identified the terms

to be used, determined corresponding references and developed a precise definition for each

attribute. The first three days of the panel consisted of free term generation, where the panelists

were unaided in the identification of the attributes present in the sample. During days 4-6 panelists

were provided with the rum flavor wheel previously developed (Chapter 5, Figure 5.1) to aid in term

generation.

After the terms and references were established, panelists then spent five sessions (1 hour each)

determining the reference intensities of each attribute. Panelists were asked to scale the references

based on a 15 point scale, where zero is no perception of a given attribute and 15 is the strongest

perception of that attribute in the rum samples. Determined reference scores can be found in Table

6.1.

Panelists then spent six days practicing scoring the rums, using the references as anchor points for

the scale, to aid in panel uniformity. On one of the days, panelists conducted individual booth

practice sessions in Bevier Hall on the University of Illinois at Urbana-Champaign campus for two

thirty-minute sessions using the Compusense five (Version 5.0: Guelph ON, Canada) data

acquisition system. Panelists were routinely provided with their scoring results from the previous day

to help identify and correct for attributes they were rating inconsistently with the rest of the group.

The panel concluded with 5 days of individual booth testing. Panelists attended two 30 minute

sessions per day, evaluating two samples at each session. Panelists were provided with their

reference tray when they arrived and encouraged to evaluate all references before proceeding into

the booth for testing. Rum samples were presented in black double old fashions glasses covered

with a petri dish and labeled with random three-digit codes. Samples were evaluated under red

lighting using the Compusense five software. Sample presentation was randomized between all

panelists. Panelists were free to leave the booth at any time to come and re-evaluate a reference.

A more detailed, day-by-day description of the DA panel can be found in Appendix O.

143

Statistical Analysis


SAS Institute Inc., Cary, NC, USA). Analysis of variance (ANOVA) was conducted on each of the

38 attributes evaluated during the DA panel to determine the presence of overall significant

differences (p<0.05) using the PROC GLM function for variations within the rums, panelists,

replications and their corresponding interactions: rum-by-panelist (Rum x P), rum-by-replication

(Rum x Rep), and replications-by-panelist (Rep x P). The calculated probabilities were compared to

significance levels α=0.05, 0.01 and 0.001. When significant Rum x P interactions existed, adjusted

F-ratios were calculated using Microsoft® Excel® 2016 (Version 16: Microsoft Corporation,

Redmond, WA) by dividing the rum mean square by the Rum x P interaction mean square and

calculating the new probability using the F.DIST function. Fisher’s least significant difference (LSD)

test was conducted on all attributes determined as significant by ANOVA.

Principal component analysis (PCA) biplots were produced using SAS and Microsoft Excel to create

a visual representation of the data to allow further examination of the relationship of the rums to

individual attributes that characterized the samples. Pearson correlations were calculated using the

same SAS software, with significance determined at α=0.05, 0.01, and 0.001. Cluster analysis was

also conducted using SAS software using the PROC CLUSTER function.


Term Generation and Lexicon Validation

The descriptive analysis panel was conducted to describe and quantify the sensory differences

perceived in the nine rum samples, two mixing and 7 premium rums, whose volatile composition

had previously been analyzed (Chapters 3 & 4). Additionally, the descriptive analysis panel provided

validation for the rum lexicon that was previously developed through the use of web-based material

(Chapter 5).

During the term generation phase of the panel, panelists were allowed to generate terms freely for

the first three days, without the aid of the flavor lexicon. This initial term generation provided terms

144

that panelists were able to identify on their own. Initially 52 terms were generated consisting of 25

aroma, 16 aroma-by-mouth, 6 mouthfeel, 1 taste, and 4 aftertaste terms. Panelists were then

provided with the flavor wheel to aid in the generation of additional terms to adequately describe the

flavor of the rums being evaluated. After the panelists were presented with the flavor wheel (Figure

5.1), an additional 17 terms were generated, including 12 aroma, 4 aroma-by-mouth and 1 aftertaste.

No additional taste or mouthfeel terms were generated. Of those additional terms generated, 14 of

the 17 terms are found on the flavor wheel. The addition of caramel aftertaste is likely not due to the

aid of the flavor wheel, but rather refinement of the panel’s ability to detect and perceive different

intensities of odorants as caramel was a term that had initially been generated in the aroma and

aroma-by-mouth modalities during the first three days of the panel.

At the conclusion of the term generation phase of the panel, 69 terms had been generated. Of the

generated terms 53 were unique attributes. The other 16 terms appeared in multiple modalities,

usually under aroma, aroma-by-mouth and aftertaste. Of the unique terms generated, 37 appear on

the flavor lexicon.

Throughout the reference refinement and scaling portions of the panel, panelists were encouraged

to reduce the number of terms to be evaluated. Terms were removed if they were no longer

perceived in the samples, panelists could not uniformly identify the attribute in the samples, or the

attribute intensity did not differ among samples. The panelists decided on a final set of 38 attributes

(Table 6.1) to describe the rum samples, consisting of 27 unique terms.

Of the final 38 terms used for evaluation, all but five terms can be found directly on the rum flavor

wheel. Of those five remaining terms, similar forms of three of the terms can be found in the

lexicon. Alcohol appears as alcoholic, and brown spice is synonymous with baking spices. Brown

sugar has both of its components (sugar and brown) found on the wheel under sugar and caramel

categories, respectively. Only two terms were not found on the wheel in any form: phenolic and

slick. When creating the rum lexicon, it was postulated that terms had been missed from the lexicon

and would need to be added as time went on. Phenolic, slick and brown sugar are three terms that

should be considered for addition to the lexicon.

Many words included in the lexicon were not selected for the descriptive analysis panel. There are

several reasons for this occurrence. First, the rums selected, primarily premium aged rums, are only a

145

selection of all the varieties of rum available. There are numerous variations of rum flavor due to

limited regulation requirements. As a result, the terms not included in the DA panel will likely be

used to describe the other categories and types of rums. Second, due to the large number of terms

generated to describe rum flavor, the terms needed to be simplified and condensed where possible

so that the panelists could consistently and accurately identify the differences among samples.

Attributes that were difficult to detect or did not differ among the rum samples were excluded from

the final evaluation. However, the large overlap between the final terms selected and the created rum

lexicon demonstrate that using web-material is a valid way to create a lexicon and can be used for

the evaluation of a variety of rums.

Descriptive Analysis Panel Results

The descriptive analysis panel was used to characterize and quantify the sensory differences between

rum samples. Panelists generated 38 terms to describe the samples. The generated terms, term

definitions, selected references, reference scores and reference preparations can be found in Table

6.1. Reference scores are an average of individual panelist’s ratings.

Analysis of variance (ANOVA) was conducted on all 38 attributes identified and rated by the

panelists, and the results are shown in Table 6.2. In general, panelists results were uniform,

indicating the high amount of training they received.

Sample replications were not a significant (p>0.05) source of variation for the majority of attributes

except citrus aroma, roasted aroma, caramel aroma-by-mouth and roasted aroma-by-mouth. The

lack of variation shows that panelists were able to consistently rate the samples between replications.

Significant variation (p<0.05) did exist for panelists for all attributes except astringent mouthfeel.

This type of variation is typical of descriptive analysis panels and is most likely attributed to panelists

only using a portion of the scale or not uniformly using the scale (Lawless & Heymann, 1999b;

Stone et al., 1974). Rep x P interactions were not significant (p>0.05) for most attributes except

caramel aroma, cherry aroma, almond aroma, cherry aftertaste, vanilla aroma-by-mouth, cherry

aroma-by-mouth, and walnut aroma-by-mouth. Lack of significant interaction indicates that the

panelists were able to agree on the intensity of attributes in the samples across replications. No

significant (p>0.05) Rum x Rep interactions were found, indicating that while the panelists may have

used the scale differently from each other, they rated the rum samples similarly across replications.

146

Across all nine rum samples, 24 of the attributes were found to be significantly (p<0.05) different.

Significant (p<0.05) Rum x P interactions existed for 17 of the attributes. These interactions indicate

that panelists did not agree on the order of attribute intensity across the rum samples. Adjusted F-

values were calculated for attributes that had significant Rum x P interactions to account for this

variation. The adjusted F-values are shown in Table 6.2. Of the adjusted F-values, five of the

attributes (caramel aroma, maple aroma, vanilla aroma, phenolic aroma, and caramel aroma-by-

mouth) were shown to be statistically (p<0.05) different among rum samples. After adjusted F-

values were calculated, twenty terms were determined to be significantly different including brown

sugar aroma, caramel aroma, maple aroma, vanilla aroma, alcohol aroma, citrus aroma, coconut

aroma, roasted aroma, smoky aroma, phenolic aroma, chocolate aroma, warming mouthfeel, slick

mouthfeel, bitter taste, brown spice aftertaste, caramel aftertaste, caramel aroma-by-mouth, maple

aroma-by-mouth, vanilla aroma-by-mouth, and coconut aroma-by-mouth.

Defining Attributes

Mean separation analysis by Fisher’s Least Significant Difference (LSD) test was performed on the

attributes that were significantly different among rum samples, and results are shown in Table 6.3.

Rums were able to be differentiated into multiple groupings for each attribute except chocolate

aroma which had two distinct groups. Significant overlap between rum groupings existed. DX had

the highest intensity score for caramel aroma, and maple aroma and both attributes were

significantly different (p<0.05) compared to the other 8 rums. DX and DR12 had the highest scores

for brown sugar aroma, vanilla aroma, chocolate aroma, caramel aftertaste, caramel aroma-by-

mouth, maple aroma-by-mouth, and vanilla aroma-by-mouth. DX, DR12, RA7, ED12, and RZ were

highest in coconut aroma and coconut aroma-by-mouth. DX, DR12, ED12, and RZ were highest in

brown spice aftertaste, while DX, DR12, RA7 and ED12 were highest in slick mouthfeel.

Interestingly, DX, DR12, AE, and BG were highest in roasted aroma while AE had the highest

smoky aroma intensity which was significantly different (p<0.05) from the other 8 rums.

Additionally, AE12, AE, and BW were highest in phenolic aroma. BW, BG, AE, AE12, RZ, and

RA7 had the highest scores for bitter taste and warming mouthfeel. BW, BG, AE, and RZ had the

highest citrus aroma. Finally, DX had the lowest alcohol aroma and was significantly different from

all other rums except DR12.

147

BW had the lowest intensity ratings for brown sugar aroma, caramel aroma, vanilla aroma, maple

aroma, chocolate aroma, slick mouthfeel, brown spice aftertaste, caramel aftertaste, caramel aroma-

by-mouth, maple aroma-by-mouth, vanilla aroma-by-mouth and coconut aroma-by-mouth. It makes

sense that BW , a clear white rum, would have the lowest intensity of these attributes as the terms

are typically associated with aging with the except coconut.

The higher number of aroma terms generated and evaluated and subsequently identified as being

significant between rums, suggests that aroma differences are easier to perceive that in-mouth

perceptions. This could potentially be a result of the high alcohol concentration.

Spider plots (Figure 6.1 and 6.2) of the results were constructed to better visualize the differences

between samples. Both the spider plots and LSD results illustrate that DX and DR12 are

significantly different from the other rums in terms of caramel aroma, vanilla aroma, chocolate

aroma, caramel aftertaste, caramel aroma-by-mouth, maple aroma-by-mouth, and vanilla aroma-by-

mouth. Interestingly, when the DX and DR12 rums are removed from the spider plot (Figure 6.3),

the remaining rums all have similar intensity profiles to each other. However, there are still attributes

that are significantly different among the rums.

A potential reason for the higher intensities of caramel, maple and vanilla attributes in the DX and

DR12 rums is that they were the only two rums found to contain ethyl vanillin (Chapter 4). Ethyl

vanillin is an artificial compound not found in nature. It is most likely added to these rums sometime

during maturation or the bottling process. While it seems peculiar for a naturally produced product

to contain a synthetic compound, the Alcohol and Tobacco Tax and Trade Bureau (TTB) lists ethyl

vanillin as one of four compounds on its limited ingredients list (Limited Ingredients, 2016). This list

identifies compounds that can be added to distilled beverages without the need to label the product

as imitation. As long as ethyl vanillin is present at concentrations less than 16 ppm, and the “total

vanillin” concentration (determined by adding the concentration of vanillin with 2.5 times the ethyl

vanillin concentration) is less than 40 ppm the final rum can still be called a natural product. The

“total vanillin” concentrations of both DX and DR12 are well under this limit, 8.75 ppm and 5.24

ppm respectively.

The Pearson correlation coefficients are given in Table 6.5. The attributes brown sugar aroma,

caramel aroma, maple aroma, vanilla aroma, coconut aroma, chocolate aroma, brown spice

148

aftertaste, caramel aftertaste, caramel aroma-by-mouth, maple aroma-by-mouth, vanilla aroma-by-

mouth, and coconut aroma-by-mouth were all positively correlated with each other (p<0.05). The

terms are all negatively correlated (p<0.05) to alcohol aroma, citrus aroma, phenolic aroma, warming

mouthfeel and bitter taste, with the exception of coconut aroma-by-mouth and warming mouthfeel

which were not correlated (p>0.05). Additionally, alcohol aroma, phenolic aroma, warming

mouthfeel and bitter taste were all positively correlated with each other. Citrus aroma was positively

correlated to alcohol aroma and warming mouthfeel. Slick mouthfeel was negatively correlated to

citrus aroma and positively correlated to caramel aroma-by-mouth and vanilla aroma-by-mouth.

Finally, chocolate aroma is negatively correlated to smoky aroma.

Principal component analysis was also conducted to help reduce the complexity of the data and gain

a better visual representation of the data (Figure 6.4) (Lawless & Heymann, 1999a). The covariance

matrix was chosen for sample evaluation since the samples were all scored by a trained panel

(Jolliffe, 2014). The first five factors have eigenvalues greater than 1 (Appendix J), and the total

variance of the products was explained in 8 factors. The first factor contained the majority of

variation between the samples (88.4%), with the second factor only containing 3.6% of the variation.

This indicates that the second factor was not significant in differentiating the rum samples. This is

also illustrated by the factor correlations (Appendix K). Principal component 1 (PC1) contrasted

samples high in brown sugar, caramel, maple, vanilla, and chocolate aromas, brown spice and

caramel aftertastes, caramel, maple, vanilla and coconut aftertastes with samples high in alcohol,

citrus, and phenolic aromas, warming mouthfeel, and bitter taste. These correlations are similar to

those indicated by the Pearson correlations. Principal component 2 (PC2) was minimally loaded with

the main distinguishing attribute being slick mouthfeel (r=0.75), as well as partially by citrus aroma

(r=-0.36) and smoky aroma (r=0.31).

Using the resulting mean intensity ratings calculated from ANOVA, cluster analysis was performed

(Figure 6.5). DX and DR12 are more closely related to each other than any of the other seven rums

since the distances between the two rums is less than half the distance between that cluster and the

other seven rums. In the grouping of 7 rums there are two distinct clusters, one containing RZ, RA7

and ED12 and a second cluster containing BW, BG, AE, and AE12 rums.

While it was expected that the Bacardi rums would be grouped more closely together, it is interesting

that they are also grouped with the Appleton Estate rums. AE and BG were grouped as most similar

149

to each other. It was hypothesized that the mixing rums (BW and BG) would be significantly

different from the premium rums. However, BW, BG, AE, and AE12 are consistently grouped

together for multiple attributes. Additionally, they tend to have the lowest intensity scores for brown

sugar aroma, caramel aroma, maple aroma, vanilla aroma, coconut aroma, brown spice aftertaste,

caramel aftertaste, caramel aroma-by-mouth, maple aroma-by-mouth, vanilla aroma-by-mouth, and

coconut aroma-by mouth.

The overall DA panel results indicate that as rums mature in the barrels, the brown sugar, caramel,

maple, and vanilla attributes across all modalities increase with age. Additionally, alcohol aroma

citrus aroma, phenolic aroma, and warming mouthfeel tend to decrease with age. However, while

overall trends are visible, clear patterns do not emerge. The rums that are clearly different from the

others are DX and DR12 with their higher caramel, maple and vanilla attributes. Even though the

remaining rums have similar profiles to each other, they can still be differentiated on a number of

attributes, demonstrating the complexity and variation present in rum products.

Comparison to Previous Sensory Studies

This was the first sensory study to evaluate premium rums and the first comprehensive study to

evaluate the in-mouth perceptions (aroma-by-mouth, taste, mouthfeel, and aftertaste) of rum flavor.

Most studies have only evaluated rum aroma (de Souza et al., 2006; Franitza et al., 2016a, 2016b;

Gómez, 2002). The only other study to study in mouth perception of rum only evaluated two rum

samples with the main focus of the study was to identify the sensory differences between rum and

cachaça (Magnani, 2009).

Comparing all six studies, alcohol aroma has been used to differentiate and characterize rum in all

studies. Additionally, one or more sweet associated terms have been used in each study: vanilla

aroma, caramel aroma or sweet aroma. Brown spice aroma is also a commonly used term ((de Souza

et al., 2006; Franitza et al., 2016a, 2016b). While Magnani (2009) also evaluated in-mouth

perceptions, only four attributes were evaluated in both studies: wood aroma-by-mouth, alcohol

aroma-by-mouth, bitter taste, and warming/burning mouthfeel. Warming/burning mouthfeel has

been labeled differently between studies, warming in the current study and residual burning in

Maganani’s study, but the term definitions were the same. This lack of overlap can be attributed to

150

difference in objectives between the studies, as Magnani was differentiating rum and cachaça

samples while the current study only differentiated rum samples.

The previous five studies evaluated 23 attributes that were not included in this study. Of those

terms, 15 are found on the rum flavor wheel. The difference in the terms selected between studies is

most likely a result of the rums being evaluated. With the huge variety of rums available, it is

reasonable that different attributes would be needed to discriminate different subsets of rum. This

demonstrates the complexity of rum flavor across products.

Little product information is typically provided in the literature, making it difficult to compare

studies and particularly the samples evaluated against each other. De Souza (2006) does identify the

brands used in her study, Bacardi was used as the rum sample, but specific product information was

not provided. Gomez (2002) is the only other author to identify the specific rums and brands

evaluated. Gomez evaluated two rums also evaluated in the current study, BW and ED12. Four

aroma attributes overlapped between the two studies, caramel, vanilla, alcohol and smoky. While

Gomez did not find any statistical difference between the two rums for these four attributes, both

studies did rate ED 12 as being higher in caramel, vanilla, and smoky aromas. Neither study found

alcohol aroma to be significantly different between the rums as indicated by their reversed order in

the two studies. The results, while limited, suggest that rum sample can be rated consistently across

studies when panelists are trained.

Evaluation of Results and Methodology

While panelists were able to differentiate rums for a number of attributes, the results demonstrate

the difficultly in evaluating distilled spirits. The large number of terms chosen by the panel to

evaluate the rum samples demonstrates the complexity of rum flavor. Sixty-nine terms were initially

generated and 38 of those were selected for the final analysis. However, only 20 of the 38 evaluated

terms were found to be significant across samples. This indicates that they panel may have needed

more training on the intensity and perception of the attributes not found to be significant. The

benefit of additional training is also indicated by the high panelist variation and Rum x P interactions

seen in the ANOVA results.

Additionally, it may be beneficial for these terms to be removed before evaluation if their intensity

differences among samples were difficult to measure. While the larger number of terms allows us to

151

better describe the total flavor profile of rum, the evaluation of so many attributes may have made it

difficult for panelists distinguish differences between samples. The large number of terms most

likely made it difficult for the panelists to remember intensities of the references when evaluating the

samples in the booth. Even though panelists were allowed to leave the booth to reevaluate reference

at any time, most did not. As a result, they may not have been able to consistently rate the product

attributes. Consolidating or reducing the terms such as brown sugar aroma, caramel aroma, and

maple aroma into one term may have aided panelists ability to focus on identifying the difference

between samples rather than trying to find the individual attributes in each sample.

Furthermore, the high ethanol concentration (40% ABV) may lead to sensory fatigue both during

training and booth testing. This fatigue may have made it difficult to accurately quantify the

differences between samples. Sensory fatigue of alcoholic beverages has been noted by other studies

(Gómez, 2002). While sample presentation was limited to two rums per session and samples order

was random, it is probable that sensory fatigue may have occurred between just these two samples

or even during the initial evaluation. However, Rum x Rep interactions only existed for caramel

aroma, indicating that if sensory fatigue occurred between samples, it did not impact the results. It is

more likely that the high alcohol concentration made it difficult to identify differences in attribute

intensities between samples and could be a contributing factor as to why so many attributes were

not significantly different among samples.

Of the 11 aroma-by-mouth terms initially identified and rated by panelists, only 4 were found to be

statistically different between rum samples; caramel, maple, vanilla, and coconut. The high amount

of panelists variation could account for the inability to differentiate the intensity of these attributes

in the samples. One reason for this could be sensory fatigue and the impact of ethanol on panelist

ability to consistently perceive aroma-by-mouth perceptions. Comparing all five modalities, panelists

had the most difficulty differentiating the aroma-by mouth attributes. More than half the attributes

were not statistically different among rums.

Additionally, panelists reluctance to make full use of the scale may have contributed to the inability

to differentiate the rums for many of the attributes. During reference scaling, panelists asked to rate

one of the samples as a 15 and base their reference and other rum scores off of that number.

However, during booth testing panelists had trouble assigning 15’s to samples indicated by the small

number of attributes receiving mean intensity scores higher than 10. The panelist's reluctance to use

152

the full scale could have resulted from improperly scored or inaccurate references, where the

reference intensity was scored too low by panelists during scaling. This could have resulted from

panelists reluctance to change the intensity of the reference even with prompting by the facilitator.

Additionally, panelists could have been reluctant to rate samples highly, expecting a sample with a

higher intensity of that attribute to be presented later.

While this panel gives significant insight into the sensory differences between premium rums and

mixing rums, it is important to note that this only encompasses a subset of available rum products.

It is expected that lower quality rums and particularly white and agricole rums will be defined and

differentiated by different sensory terms as indicated from the most frequent terms used to describe

those products (Chapter 5). Additionally, DA panels quantify the sensory differences among samples

and may not be the best at characterizing the overall flavor of rums. To better evaluate the entirely

of rum flavor it may be best to perform a preliminary sensory evaluation and sorting of a larger

variety of rum samples using techniques such as Napping to identify similar types of rums and then

select one sample from each category to be included in a descriptive analysis panel or aroma profile

analysis. This would provide greater insight into the differences between rum across the class.

153


Table 6.1 List of final attributes, definitions, references, reference scores, and reference preparations determined for the descriptive analysis panel on rum

Modality Attribute Definition Reference Reference Score Preparation

Aro

ma

Sugar & Spices

brown sugar aroma of brown sugar dark brown

sugar 13.5 1 packed teaspoon in 2 oz. cup

caramel aroma of caramelized sugar Smucker's

caramel syrup 14 10 g in 2 oz. cup

maple aroma of maple syrup maple extract 11.5 1 3/4 teaspoon in 500 mL volumetric flask, dilute to

volume, 10 mL in 2 oz. cup

Baking Spices

brown spice aroma of nutmeg ground nutmeg

14 1g in 600 mL of water, stir 5 minutes, 10 mL in 2 oz.

cup

vanilla aroma of vanilla extract natural vanilla

extract 10.5

1/4 teaspoon in 500 mL volumetric flask, dilute to volume, 10 mL in 2 oz. cup

Alcohol alcohol aroma associated with 40%

or greater alcohol 190 Proof Ethanol

11.5 10 oz. in 2 oz. cup

Fruit

cherry aroma of cooked cherries cherry pie

filling 14.5 1 cherries and syrup to 5 g in 4 oz. cup

citrus aroma of lime zest lime peel 12 soak 0.5 g in 200 mL of hot water for 5 minutes, 10 mL

in 2oz cup

coconut aroma of toasted coconut toasted

coconut flakes 14 0.1 g in 4 oz. cup

dried fruit aroma of prunes prunes 14.5 0.8 g (~1/8) prune in a 2 oz. cup

Nutty

almond aroma of almonds almonds 13.5 1 whole almond in 4 oz. cup

walnut aroma of walnuts chopped walnuts

12.5 1 g in 2 oz. cup

Woody/ Roasted

roasted aroma of medium roasted

malted barley brown roasted

barley 15 0.2 g in a 2 oz. cup

smoky aroma associated with a

hardwood smoked products smoked bacon 15 0.02 g piece of bacon, just muscle tissue, in a 4 oz. cup

woody aroma of a wood barrel oak wood

chips 11 1 g in 2 oz. cup

Medicinal phenolic aroma of a Band-Aid Band-Aid 13 1 unwrapped Band-Aid in a 2 oz. cup

Dairy butter aroma of melted butter melted butter 13 2 TBS of better melted over low heat, 0.2 g in a 4 oz.

cup

chocolate aroma of dark chocolate baker’s

chocolate 13 0.01 g of shaved chocolate in a 2 oz. cup

154

Table 6.1 (cont.) List of final attributes, definitions, references, reference scores, and reference preparations determined for the descriptive analysis panel on rum.


Mouthfeel

astringent a drying sensation in the

mouth associated with a high tannin wine

over-brewed green tea

8.5 steep 1 tea bag in 300 mL of boiling water for 5

minutes, place 15 mL in a 2 oz. cup

slick a smooth tongue coating glycerin 9.5 20 g of glycerin + 60 g water, ~10 g in a 2 oz. cup

warming

the increase in temperature perception in the mouth as a

result of alcohol concentration

190 Proof Ethanol (1:2

dilution) 14 1:2 dilution, 10 mL in 2 oz. cup (alcohol)

Taste bitter bitter taste of caffeine

solution caffeine solution

10 1 g caffeine in 500 mL of water, 15 mL in each cup

Aftertaste

bitter aftertaste associated with a

caffeine solution caffeine solution

11.5 1 g caffeine in 500 mL of water, 15 mL in each cup

brown spice aftertaste associated with

brown spices nutmeg 14

1 g in 600 mL of water, stir 5 minutes, 10 mL in 2 oz. cup

caramel aftertaste associated with

caramel Smucker's

caramel syrup 13

caramel solution, 10 g of caramel dissolved in 200 mL of water ~10 mL in a 2 oz. cup

cherry aftertaste associated with

cooked cherries cherry pie

filling 13 1 cherry and syrup to 5 g in 4 oz. cup

coffee aftertaste associated with

coffee dark roast

coffee 12

1 1/2 teaspoons of coffee, in teabag, in 300 mL of boiling water for 5 minutes

155

Table 6.1 (cont.) List of final attributes, definitions, references, reference scores, and reference preparations determined for the descriptive analysis panel

on rum.


Aro

ma b

y M

ou

th

Sugar

caramel aroma-by-mouth of caramelized sugar

caramel syrup solution

10 caramel solution, 10g of caramel dissolved in 200mL of

water ~10mL in a 2 oz. cup

maple aroma-by-mouth of maple

syrup maple extract 13

1 3/4 teaspoon in 500mL volumetric flask, dilute to volume, 10 mL in 2oz cup

Baking Spices

brown spice aroma-by-mouth of nutmeg ground nutmeg

15 1g in 600mL of water, stir 5 minutes, 10mL in 2 oz.

cup

vanilla aroma-by-mouth of vanilla

extract natural vanilla

extract 11

1/4 teaspoon in 500mL volumetric flask, dilute to volume, 10 mL in 2oz cup

Alcohol alcohol aroma-by-mouth associated with 40% or greater alcohol

Everclear 14 10 oz. in 2 oz. cup

Fruit

cherry aroma-by-mouth of cooked

cherries cherry pie

filling 13 1 cherries and syrup to 5g in 4 oz. cup

coconut aroma-by-mouth of toasted

coconut toasted

coconut flakes 14 1g in 2 oz. cup

Nutty walnut aroma-by-mouth of walnuts chopped walnuts

13.5 3g in 2 oz. cup

Woody/ Roasted

roasted aroma-by-mouth of medium

roasted malted barley brown roasted

barley 15 0.2 g in a 2oz cup

smoky aroma-by-mouth associated with a hardwood smoked

products liquid smoke 14.5

0.5g in 1000mL volumetric flask, dilute to volume, 10 mL in 2oz cup

woody aroma-by-mouth of a wood

barrel oak wood

chips 8 1g in 2 oz. cup

156

Table 6.2 Analysis of variance (ANOVA) F-ratios for sensory attributes rated for nine rum samples+

Modality Attribute Rum Panelist Rep Rum x P† Rum x Rep† Rep x P†

Adjusted Sample F

Aroma

brown sugar 7.44*** 5.53*** 0.11 1.30 0.78 1.67

caramel 31.87*** 9.07*** 2.88 2.01** 2.15* 2.86* 15.82***

maple 13.15*** 7.73*** 0.00 1.71* 1.31 1.80 7.67**

brown spice 1.48 6.60*** 2.54 2.45*** 1.26 2.04 0.61

vanilla 11.64*** 6.00*** 0.14 1.89** 2.09 0.92 6.16*

alcohol 3.62** 5.82*** 0.04 1.05 0.77 0.87

cherry 1.77 19.85*** 0.48 2.05** 1.24 2.43* 0.86

citrus 3.52*** 16.91*** 4.41* 1.09 0.68 1.26

coconut 4.15*** 6.11*** 0.10 1.33 0.96 0.74

dried fruit 1.26 8.54*** 2.57 2.25** 1.28 1.88 0.67

almond 5.16*** 16.84*** 0.01 2.44*** 0.99 2.42* 2.12

walnut 1.49 5.56*** 1.26 0.91 0.31 1.08

roasted 2.40* 7.37*** 4.30* 0.98 0.73 1.48

smoky 3.83** 2.51* 0.70 1.06 1.75 0.71

woody 0.99 4.88*** 0.79 1.55 1.31 1.04

phenolic 9.00*** 9.14*** 0.69 2.22** 1.48 0.49 4.06**

butter 7.44*** 7.30*** 0.73 2.14** 1.46 0.92 3.48

chocolate 4.65*** 5.79*** 0.04 1.27 0.44 1.36

Mouthfeel

astringent 1.48 1.03 3.22 1.38 0.25 1.31

warming 2.65* 5.29*** 0.03 0.92 0.74 1.29

slick 3.18** 9.34*** 0.51 1.20 0.51 0.53

Taste bitter 3.47** 3.05** 0.04 1.22 1.67 1.31

Aftertaste

brown spice 3.73** 5.54*** 2.52 2.33*** 0.85 0.57 1.60

caramel 23.19*** 5.28*** 1.89* 1.40 1.59 1.35

cherry 0.77 14.10*** 0.86 2.52*** 1.18 2.71* 0.30

coffee 1.90 5.35*** 0.60 1.09 0.79 1.77

Aroma by Mouth

caramel 29.65*** 14.34*** 5.54 2.05** 1.42 1.22 14.48***

maple 8.02*** 5.84*** 0.50 1.20 1.08 1.04

brown spice 3.32** 12.07*** 0.67 3.10*** 1.79 1.84 1.07

vanilla 12.22*** 12.59*** 1.44 1.31 0.88 2.40*

alcohol 2.65* 4.68*** 0.32 1.18 2.01 0.74

cherry 0.72 17.17*** 0.02 2.02** 0.59 4.61*** 0.26

coconut 7.50*** 5.35*** 1.95 2.46*** 0.99 1.11 3.05

walnut 1.39 31.99*** 0.01 1.71* 1.08 3.03* 0.82

roasted 0.88 9.02*** 8.65** 1.28 1.45 1.06

smoky 1.37 7.86*** 1.03 1.18 1.18 1.43

woody 0.93 6.13*** 0.21 1.89** 0.86 0.80 0.49 +F-rations are shown as a source of variation. *, **, *** stand for significance at p<0.05, p<0.01, and p<0.001 respectively. †Rum x P, Rep x P, and Rum x Rep represent the interaction between rum samples and panelists, replications and panelists, and rum samples and replications, respectively.

157

Table 6.3 Mean intensity rating for significant aroma, mouthfeel, taste, aftertaste and aroma-by-mouth attributes of the nine rum samples†

Modality Attribute BW* BG* AE* RA7* AE12* DR12* ED12* RZ* DX*

Aroma

brown sugar 3.06D 4.69C,D 4.69C,D 6.71B,C 5.88C 8.38A,B 5.75C 5.63C 9.94A

caramel 3.06E 5.31C,D 4.25D,E 5.88C,D 4.44D,E 11.19B 4.88C,D 6.44C 13.75A

maple 3.21D 4.81C,D 4.88C,D 6.38B,C 5.75C 7.81B 6.00B,C 5.63C 12.31A

vanilla 4.44C 6.13B,C 5.75B,C 7.19B 5.43B,C 11.06A 6.81 B 6.81 B 11.81A

alcohol 9.69A 9.75A 9.06A,B 9.00A,B 8.88A,B 7.25B,C 8.38A,B 8.38A,B 5.56C

citrus 4.50A,B 4.81A 3.31A,B,C 2.94B,C,D 2.94B,C,D 2.31C,D 2.28C,D 4.13A,B 1.50D

coconut 2.50C,D 2.94B,C,D 2.69C,D 4.94A,B 2.19D 5.06A 5.75A 4.31A,B,C 6.06A

roasted 2.69B,C 3.44A,B 3.31A,B 2.88B 1.19C 4.50A 2.44B,C 2.63B,C 3.00A,B

smoky 2.19B,C 3.38B 5.44A 2.63B,C 2.06B,C 1.00C 2.75B,C 2.00B,C 1.25C

phenolic 7.50A,B 5.63B 8.19A 5.38B 8.06A 2.69C 5.50B 5.81B 1.31C

chocolate 1.50B 1.63B 1.13B 2.25B 2.31B 4.06A 2.31B 2.44B 4.69A

Mouthfeel warming 8.56A,B,C 9.25A,B 8.81A,B,C 8.69A,B,C 9.38A 7.69B,C,D 7.63C,D 8.56A,B,C 6.50D

slick 4.44D 5.81C,D 7.18A,B,C 6.38A,B,C 6.13B,C,D 8.00A 7.81A,B 6.00C,D 6.69A,B,C

Taste bitter 8.06A,B 7.81A,B,C 8.31A,B 7.75A,B,C 8.94A 6.50C,D 6.94B,C,D 8.13A,B 5.75D

Aftertaste brown spice 4.75D 6.31B,C 5.50C,D 6.56B,C 6.13B,C,D 7.56A,B 7.56A,B 6.88A,B,C 8.19A

caramel 1.50D 2.88C,D 2.75C,D 4.13C 4.00C 10.31A 6.13B 4.38B,C 10.75A

Aroma By Mouth

caramel 3.00E 3.88D,E 4.56D,E 6.88B 5.19B,C 12.19A 7.25B 6.63B,C 11.13A

maple 2.75C 3.88B,C 4.00B,C 5.31B 5.13B 8.19A 6.00B 5.69B 9.69A

vanilla 4.31D 5.56C,D 6.81B,C 7.56B 6.88B,C 11.31A 8.56B 7.00B,C 11.00A

coconut 1.31D 3.38B,C 2.56C,D 5.06A 2.63C,D 5.25A 4.81A,B 4.31A,B 5.56A

†Superscripts of the same letter within an attribute indicate no significant difference by Fisher’s Least Significant Difference (LSD) test at α=0.05 *”BW” is Bacardi White, “BG” is Bacardi Gold, “AE” is Appleton Estate V/X, “RA7” is Ron Abuelo 7 year, “AE12” is Appleton Estate 12 year, “DR12” is Diplomatico Reserva Exclusica, “ED12” is El Dorado 12 year, “RZ” is Ron Zacapa Centurio, “DX” is Dictador XO Insolent

158

Figure 6.1 Spider plots of mean significant attribute intensities for all nine rum samples. “BW” is Bacardi White, “BG” is Bacardi Gold, “AE” is Appleton Estate V/X, “RA7” is Ron Abuelo 7 year, “AE12” is Appleton Estate 12 year, “DR12” is Diplomatico Reserva Exclusica, “ED12” is El Dorado 12 year, “RZ” is Ron Zacapa Centurio, “DX” is Dictador XO Insolent. “A” is aroma, “M” is mouthfeel, “T” is taste, “AT” is aftertaste, “ABM” is aroma-by-mouth, “BSp” is brown spice, “BSu” is brown sugar, “Ca” is caramel, “M” is maple, “V” is vanilla, “Co” is coconut, “Cit” is citrus, “P” is phenolic, “B” is Bitter, “A” is alcohol, “W” is warming, “Sl” is slick, “Sm” is smoky, “R” is roasted, and “Ch” is chocolate.

02468

101214BSp_AT

BSu_ACa_A

Ca_ABM

Ca_AT

M_A

M_ABM

V_A

V_ABMCo_A

Co_ABMCit_A

P_A

B_T

A_A

W_M

Sl_M

Sm_A

R_ACh_ABW

02468

101214BSp_AT

BSu_ACa_A

Ca_ABM

Ca_AT

M_A

M_ABM

V_A

V_ABMCo_A

Co_ABMCit_A

P_A

B_T

A_A

W_M

Sl_M

Sm_A

R_ACh_ABG

02468

101214BSp_AT

BSu_ACa_A

Ca_ABM

Ca_AT

M_A

M_ABM

V_A

V_ABMCo_A

Co_ABMCit_A

P_A

B_T

A_A

W_M

Sl_M

Sm_A

R_ACh_AAE

02468

101214BSp_AT

BSu_ACa_A

Ca_ABM

Ca_AT

M_A

M_ABM

V_A

V_ABMCo_A

Co_ABMCit_A

P_A

B_T

A_A

W_M

Sl_M

Sm_A

R_ACh_ARA7

02468

101214BSp_AT

BSu_ACa_A

Ca_ABM

Ca_AT

M_A

M_ABM

V_A

V_ABMCo_A

Co_ABMCit_A

P_A

B_T

A_A

W_M

Sl_M

Sm_A

R_ACh_AAE12

02468

101214BSp_AT

BSu_ACa_A

Ca_ABM

Ca_AT

M_A

M_ABM

V_A

V_ABMCo_A

Co_ABMCit_A

P_A

B_T

A_A

W_M

Sl_M

Sm_A

R_ACh_A

DR12

02468

101214BSp_AT

BSu_ACa_A

Ca_ABM

Ca_AT

M_A

M_ABM

V_A

V_ABMCo_A

Co_ABMCit_A

P_A

B_T

A_A

W_M

Sl_M

Sm_A

R_ACh_AED12

02468

101214BSp_AT

BSu_ACa_A

Ca_ABM

Ca_AT

M_A

M_ABM

V_A

V_ABMCo_A

Co_ABMCit_A

P_A

B_T

A_A

W_M

Sl_M

Sm_A

R_ACh_ARZ

02468

101214BSp_AT

BSu_ACa_A

Ca_ABM

Ca_AT

M_A

M_ABM

V_A

V_ABMCo_A

Co_ABMCit_A

P_A

B_T

A_A

W_M

Sl_M

Sm_A

R_ACh_ADX

159

Figure 6.2 Overlaid spider plot of mean significant attribute intensities for all nine rum samples. “BW” is Bacardi White, “BG” is Bacardi Gold, “AE” is Appleton Estate V/X, “RA7” is Ron Abuelo 7 year, “AE12” is Appleton Estate 12 year, “DR12” is Diplomatico Reserva Exclusica, “ED12” is El Dorado 12 year, “RZ” is Ron Zacapa Centurio, “DX” is Dictador XO Insolent. A is aroma, M is mouthfeel, T is taste, AT is aftertaste, ABM is aroma-by-mouth

0

2

4

6

8

10

12

14

BrownSpice_AT

BrownSugar_A

Caramel_A

Caramel_ABM

Caramel_AT

Maple_A

Maple_ABM

Vanilla_A

Vanilla_ABM

Coconut_A

Coconut_ABM

Citrus_A

Phenolic_A

Bitter_T

Alcohol_A

Warming_M

Slick_M

Smoky_A

Roasted_A

Chocolate_A

AE AE12 BG BW DR12 DX ED12 RA7 RZ

160

Figure 6.3 Spider plot of mean significant attribute intensities for seven rum samples, excluding Diplomatica Reverva 12 year and Dictador Insolent “BW” is Bacardi White, “BG” is Bacardi Gold, “AE” is Appleton Estate V/X, “RA7” is Ron Abuelo 7 year, “AE12” is Appleton Estate 12 year, “DR12” is Diplomatico Reserva Exclusica, “ED12” is El Dorado 12 year, “RZ” is Ron Zacapa Centurio, “DX” is Dictador XO Insolent. A is aroma, M is mouthfeel, T is taste, AT is aftertaste, ABM is aroma-by-mouth

0.0

2.0

4.0

6.0

8.0

10.0

BrownSpice_AT

BrownSugar_A

Caramel_A

Caramel_ABM

Caramel_AT

Maple_A

Maple_ABM

Vanilla_A

Vanilla_ABM

Coconut_A

Coconut_ABM

Citrus_A

Phenolic_A

Bitter_T

Alcohol_A

Warming_M

Slick_M

Smoky_A

Roasted_A

Chocolate_A

AE AE12 BG BW ED12 RA7 RZ

161

Figure 6.4 Principal component analysis biplot of significant attributes present on principal component 1 (PC1) and 2 (PC2) by the covariance matrix of mean significant attribute intensity rating across all nine rum samples “BW” is Bacardi White, “BG” is Bacardi Gold, “AE” is Appleton Estate V/X, “RA7” is Ron Abuelo 7 year, “AE12” is Appleton Estate 12 year, “DR12” is Diplomatico Reserva Exclusica, “ED12” is El Dorado 12 year, “RZ” is Ron Zacapa Centurio, “DX” is Dictador XO Insolent. A is aroma, M is mouthfeel, T is taste, AT is aftertaste, ABM is aroma-by-mouth.

AE

AE12

BG

BW

DR12

DX

ED12

RA7

RZ

BrownSugar_A

Caramel_A

Maple_AVanilla_A

Alcohol_A

Citrus_A

Coconut_A

Roasted_A

Smoky_A

Phenolic_A

Chocolate_A

Warming_M

Slick_M

Bitter_T

BrownSpice_AT

Caramel_ATCaramel_ABM

Maple_ABM

Vanilla_ABMCoconut_ABM

-1 0 1

-1

0

1

-2

0

2

-2 0 2

PC

2 (

3.6

1%)

PC1 (88.36%)

Variables (axes PC1 and PC2: 91.97%)

162

Figure 6.5 Cluster analysis constructed using mean significant attribute intensities for all nine rums “BW” is Bacardi White, “BG” is Bacardi Gold, “AE” is Appleton Estate V/X, “RA7” is Ron Abuelo 7 year, “AE12” is Appleton Estate 12 year, “DR12” is Diplomatico Reserva Exclusica, “ED12” is El Dorado 12 year, “RZ” is Ron Zacapa Centurio, “DX” is Dictador XO Insolent

163

Table 6.4 Pearson correlation coefficients for significant attributes for all nine rum samples

A is aroma, M is mouthfeel, T is taste, AT is aftertaste, ABM is aroma-by-mouth. *, **, *** stand for significance at p<0.05, p<0.01, and p<0.001 respectively.

Attributes BrownSugar

A Caramel

A Maple

A Vanilla

A Alcohol

A Citrus

A Coconut

A Roasted

A Smoky

A Phenolic

A

BrownSugar_A 1.000

Caramel_A 0.932** 1.000

Maple_A 0.955*** 0.932** 1.000

Vanilla_A 0.945** 0.983*** 0.910** 1.000

Alcohol_A -0.916** -0.935** -0.956*** -0.921** 1.000

Citrus_A -0.842** -0.681* -0.812** -0.741* 0.820** 1.000

Coconut_A 0.752* 0.701* 0.729* 0.773* -0.743* -0.700* 1.000

Roasted_A 0.273 0.466 0.194 0.525 -0.237 -0.048 0.312 1.000

Smoky_A -0.565 -0.598 -0.511 -0.570 0.567 0.364 -0.465 -0.001 1.000

Phenolic_A -0.867** -0.940** -0.861** -0.950*** 0.847** 0.606 -0.827** -0.522 0.635 1.000

Chocolate_A 0.936** 0.953*** 0.913** 0.945** -0.940** -0.752* 0.726* 0.280 -0.772* -0.908**

Warming_M -0.730* -0.794* -0.811** -0.813** 0.885** 0.740* -0.838** -0.325 0.488 0.827**

Slick_M 0.565 0.436 0.439 0.565 -0.490 -0.705* 0.582 0.377 0.026 -0.414

Bitter_T -0.764* -0.845** -0.801** -0.881** 0.821** 0.662 -0.850** -0.535 0.487 0.933**

BrownSpice_AT 0.833** 0.759* 0.827* 0.790* -0.803** -0.711* 0.878** 0.130 -0.482 -0.812**

Caramel_AT 0.933** 0.936** 0.892** 0.966*** -0.936** -0.823** 0.783* 0.392 -0.628 -0.902**

Caramel_ABM 0.928** 0.912** 0.841** 0.960*** -0.894** -0.805** 0.806** 0.448 -0.623 -0.882**

Maple_ABM 0.969*** 0.942** 0.945** 0.954*** -0.966*** -0.838** 0.800** 0.271 -0.619 -0.888**

Vanilla_ABM 0.934** 0.880** 0.858** 0.938** -0.897** -0.881** 0.792* 0.397 -0.497 -0.830**

Coconut_ABM 0.844** 0.753* 0.756* 0.821** -0.718* -0.674* 0.922 0.351 -0.450 -0.835**

164

Table 6.4 (cont.) Pearson correlation coefficients for significant attributes for all nine rum samples

A is aroma, M is mouthfeel, T is taste, AT is aftertaste, ABM is aroma-by-mouth. *, **, *** stand for significance at p<0.05, p<0.01, and p<0.001 respectively.

Attributes Chocolate

A Warming

M Slick

M Bitter

T BrownSpice

AT Caramel

AT Caramel

ABM Maple ABM

Vanilla ABM

Coconut ABM

BrownSugar_A

Caramel_A

Maple_A

Vanilla_A

Alcohol_A

Citrus_A

Coconut_A

Roasted_A

Smoky_A

Phenolic_A

Chocolate_A 1.000

Warming_M -0.802** 1.000

Slick_M 0.424 -0.444 1.000

Bitter_T -0.813** 0.933** -0.487 1.000

BrownSpice_AT 0.789* -0.729* 0.608 -0.757* 1.000

Caramel_AT 0.961*** -0.838** 0.642 -0.867** 0.830** 1.000

Caramel_ABM 0.935** -0.788* 0.672* -0.819** 0.798** 0.976*** 1.000

Maple_ABM 0.966*** -0.823** 0.596 -0.818** 0.881** 0.977*** 0.958*** 1.000

Vanilla_ABM 0.895** -0.782* 0.775* -0.804** 0.821** 0.972*** 0.978*** 0.957*** 1.000

Coconut_ABM 0.752* -0.659 0.654 -0.745* 0.916** 0.803** 0.849** 0.839** 0.839** 1.000

165

6.6 References

Cairncross, W. E., & Sjöström, L. B. (1950). Flavor Profile - a new approach to flavor problems.

Food Technology, 4, 308–311.

Caul, J. F. (1957). The Profile Method of Flavor Analysis. Advances in Food Research, 7, 1–40.



488. https://doi.org/10.1021/jf0511190









Ickes, C. M., Lee, S.-Y., & Cadwallader, K. R. (2017). Novel Creation of a Rum Flavor Lexicon

Through the Use of Web-Based Material. Journal of Food Science, 82(5), 1213–1223.

https://doi.org/10.1111/1750-3841.13707

Jolliffe, I. (2014). Principal Component Analysis. In Wiley StatsRef: Statistics Reference Online (pp. 1–5).

Lawless, H. T., & Heymann, H. (1999a). Data Relationships and Multivariate Applications. In Sensory

Evaluation of Food: Principles and Practices (pp. 585–601). New York: Springer Science+Business

Media, LLC.

Lawless, H. T., & Heymann, H. (1999b). Descriptive analysis. In Sensory Evaluation of Food: Principles

and Practices (pp. 227–257). New York: Springer Science & Buiness Media, LLC.

https://doi.org/10.1007/978-1-4419-6488-5


166




Meilgaard, M., Civille, G. V., & Carr, B. T. (2007). The Spectrum Descriptive Analysis Method. In

Sensory Evaluation Techniques (Fourth Edi, pp. 189–254). Boca Raton, FL: CRC Press/Taylor &

Francis.

Muñoz, A. M., & Civille, G. V. (1992). The Spectrum Descriptive Analysis Method. In R. C.

Hootman (Ed.), Manual on Descriptive Analysis Testing for Sensory Evaluation (pp. 22–34). Baltimore,

MD: American Society for Testing and Materials.

Powers, J. J. (1988). Current Practices and Application of Descriptive Methods. In J. R. Piggott

(Ed.), Sensory Analysis of Foods (2nd ed., pp. 187–266). New York: Elsevier Applied Science.

Stone, H. (1992). Quantitative Descriptive Analysis. In R. C. Hootman (Ed.), Manual on Descriptive

Analysis Testing for Sensory Evaluation (pp. 15–21). Baltimore, MD: American Society for Testing

and Materials.

Stone, H., & Sidel, J. L. (1993). Descriptive Analysis. In S. L. Taylor (Ed.), Sensory Evaluation Practices

(2nd ed., pp. 202–242). San Diego: Academic Press Inc.

Stone, H., Sidel, J. L., Oliver, S., Woolsey, A., & Singleton, R. (1974). Sensory Evaluation by

Quantitative Descriptive Analysis. Food Technology, 8(1), 24–32.

https://doi.org/10.1002/9780470385036.ch1f

Williams, A. A., & Langron, S. P. (1984). The Use of Free choice Profiling for the Evaluation of

Commercial Ports. Journal of the Science of Food and Agriculture, 35(5), 558–568.

https://doi.org/10.1002/jsfa.2740350513

Zook, K. L., & Pearce, J. H. (1988). Quantitative Descriptive Analysis. In H. Moskowitz (Ed.),

Applied Sensory Analysis of Foods: Volume 1 (pp. 43–72). Boca Raton, FL: CRC Press, Inc.

167

Chapter 7: Effect of Ethanol Concentration on the Flavor Perception

of Rum

7.1 Abstract

This is the first sensory study to evaluate the effects of ethanol concentration on the flavor

perception of distilled spirits. Dilution series of two rum samples were evaluated to gain insight into

the effects of ethanol concentration on the flavor perception of distilled spirits. Rum samples were

diluted 1:2 with either pure water to obtain a final sample of 20% ABV or with a 40% ABV solution

to account for the flavor dilution effect but keep the ethanol concentration the same as the original

samples. A descriptive analysis panel was conducted on both dilution series. Twenty-three attributes

were evaluated consisting of eight aroma, four mouthfeel, two taste, five aftertaste and four aroma-

by-mouth terms. The results indicated that the original rum samples had the highest intensity for all

attributes and the dilution with 40%ABV had the lowest perception except for silky mouthfeel. The

flavor profiles of the original sample and the dilutions with water were very similar, with the water

dilutions generally having lower attribute intensities. In contrast, the dilution with 40% ABV had

significantly different flavor profiles.

7.2 Introduction

Ethanol is arguably the most important component of alcoholic beverages, particularly distilled

spirits. Alcoholic beverages span a wide range of alcohol concentrations from beers (3% to 10%

ABV) to distilled spirits (usually 40% ABV). Furthermore, ethanol concentration plays an important

role in the flavor perception of alcoholic beverages. Molecular structure, flavor partitioning, and

sensory perceptions have all been shown to be affected as a function of ethanol concentration.

In terms of the physicochemical properties of ethanol, the water-ethanol structure greatly changes as

the system goes from 100% aqueous to 100% ethanolic (Cipiciani, Onori, & Savelli, 1988; Conner,

Birkmyre, Paterson, & Piggott, 1998; D’Angelo, Onori, & Santucci, 1994a, 1994b; Franks & Ives,

1966; Onori & Santucci, 1996; Parke & Birch, 1999; Petrillo, Onori, & Sacchetti, 1989; Wakisaka,

168

Komatsu, & Usui, 2001). In 100% aqueous systems, the water molecules exist in short-range

hydrogen bonded structures that are highly fluid, continuously breaking and reforming hydrogen

bonds (Franks & Ives, 1966). As the ethanol concentration initially increases, the ethanol molecules

are mono-dispersed within the aqueous matrix (Conner et al., 1998; D’Angelo et al., 1994a, 1994b).

This behavior continues until the ethanol concentration reaches approximately 15% alcohol-by-

volume (ABV). Above this concentration, ethanol molecules begin to form aggregates or micelles

(Cipiciani et al., 1988; D’Angelo et al., 1994b; Onori & Santucci, 1996). Once the solution reaches an

ethanol concentration of 57% ABV, the final structural shift in the water/ethanol matrix occurs

where the water molecules become mono-dispersed within the ethanolic matrix (D’Angelo et al.,

1994b).

Previous research on how ethanol concentration affects aroma and sensory perception in alcoholic

beverages has mainly approached the problem from an analytical perspective. Additionally, studies

have mainly focused on how ethanol affects the headspace concentration of volatiles in static

systems. As the ethanol concentration of solutions increases, the headspace concentration has been

shown to decrease (Athès, Pena-Lillo, Bernard, Perez-Correa, & Souchon, 2004; Aznar, Tsachaki,

Linforth, Ferreira, & Taylor, 2004; Boothroyd, Linforth, & Cook, 2012; Conner et al., 1998;

Escalona-Buendia, Piggott, Conner, & Paterson, 1998; Tsachaki, Aznar, Linforth, & Taylor, 2006).

The decrease in headspace concentration is typically attributed to an increase in the solubility of

aroma compounds as the ethanol concentration increases. While overall decreases in headspace

concentration occur, it is important to note volatile compounds are affected differently by the

changes in ethanol concentration (Aznar et al., 2004; Boothroyd et al., 2012; Tsachaki et al., 2006).

While these studies demonstrate that ethanol can affect the release of aroma compounds in alcoholic

beverages, they are not representative of real life systems.

When beverages are consumed, they are not evaluated in equilibrium systems. The glasses are

constantly open to the air, allowing solvent evaporation, and exposure to air currents in the room.

Additionally, the glass is stirred, either slowly as it is moved for consumption, or deliberately. As a

result, the study of dynamic systems allows for better insight into how ethanol concentration may

affect sensory perception during the act of consumption.

Evaluation of dynamic systems has mainly been done using atmospheric pressure chemical

ionization (APCI)-MS (Taylor et al., 2010; Tsachaki et al., 2008; Tsachaki, Linforth, & Taylor, 2005).

169

Results from these studies contradict those of the static systems, showing that the headspace

concentration increases above ethanolic systems in comparison to aqueous systems under dynamic

conditions. This phenomenon is attributed to a combination of the Marangoni effect and Rayleigh-

Bénard convection (Taylor et al., 2010; Tsachaki et al., 2006, 2005). Essentially, the bulk solution is

stirred through differences in surface tension and density respectively as caused by the evaporation

of ethanol (Marangoni, 1865; Rayleigh, 1916). The stirring moves more ethanol and flavor molecules

to the surface of the solution for continued evaporation. The results of these studies suggest that

alcoholic beverages with higher alcohol concentrations will have increased sensory attribute intensity

due to the higher concentration of aroma compounds in the headspace above the samples.

Overall, sensory research exploring the effects of ethanol concentration on the perception of

alcoholic beverages is lacking. The majority of studies have evaluated ethanol’s effect on wine or

wine model systems (Escudero, Campo, Fariña, Cacho, & Ferreira, 2007; Goldner, Zamora, Di Leo

Lira, Gianninoto, & Bandoni, 2009; Guth, 1998; Jones, Gawel, Francis, & Waters, 2008; King,

Dunn, & Heymann, 2013; Le Berre, Atanasova, Langlois, Etiévant, & Thomas-Danguin, 2007).

While these studies demonstrate that ethanol concentration can have an impact on flavor perception

in alcohol, the alcohol content of wine is very different from distilled spirits.

To date, the only sensory study to include distilled spirits was done by in the 1970’s (Williams, 1972).

Williams dealcoholized five different alcoholic beverages and observed the sensory changed

compared to the original beverage. The study showed that de-ethanolized whiskey perceived as drier

and had less bite than the full alcohol sample. No sensory research had been conducted to provide

insights into how ethanol’s concentration may affect flavor perception during the consumption of

distilled spirits.

While some consumers will only drink distilled spirits neat, others typically dilute their drink before

consumption. A common practice is to consume spirits on the rocks, where the ice both cools and

dilutes the beverage. Others insist that a small splash of water is needed to open up the flavor. Still

others state that distilled spirits need to be diluted to ~23% ABV to get the perception of the

beverage. This has been the traditional practice in the whiskey industry for years, and the common

reason given for this practice is to reduce the pungency of the alcohol (Smith & Roskrow, 2012).

Descriptive analysis panels allow researchers to identify and quantify the sensory differences

between products (Lawless & Heymann, 1999b; Meilgaard, Civille, & Carr, 2007; Stone & Sidel,

170

1993). Panelists develop terms and corresponding definitions and references for the sensory

attributes they perceive in the sample for the modalities of aroma, aroma-by-mouth, taste,

mouthfeel, and aftertaste. Panelists rate the determined references and then use those as anchor

points on the scale when evaluating the samples. Panelists finally rate the samples in individual

booths for all the attributes that were previously identified. The collected data is then subjected to

statistical analysis to evaluate the samples using methods such as analysis of variance, and principal

component analysis.

The goal of this study was to evaluate using a descriptive analysis panel, the effects of ethanol

concentration on the flavor perception of distilled spirits using rum samples by comparing the

original sample to a 1:2 dilution with water and a 1:2 dilution with 40% ethanol. It is hypothesized

that changes in ethanol concentration will significantly affect the aroma of rum samples. Samples

with lower ethanol concentrations are expected to have lower perceived solvent and pungent

characteristics. The 40% ABV dilution was expected to be most similar to the original rum sample.


Materials

Two commercially available rums, Ron Abuelo: Añejo 7 years (RA7), and Diplomatica Reserva

Exclusiva (DR12), were purchased at a local liquor store (Champaign, IL). Both rums had a reported

ethanol concentration of 40% alcohol by volume (ABV). Mention of the brand name of these rums

does not imply any research contact or sponsorship and is not for advertisement or endorsement

purposes.

Two dilutions of each rum were prepared, a 1:2 (v/v) dilution with 40% ethanol and a 1:2 (v/v)

dilution with pure water (Ice Mountian 100% Natural Spring Water, Nestlé Waters North America,

Stamford, CT). The 40% ethanol solution was prepared by diluting 190 Proof (95%ABV) ethanol

(USP Grade, Decaon Labs, Inc., King of Prussia, PA) with pure water. Prior to testing, 20 mL of

each sample was measured into a black double old-fashion glass (Threshold, Target, USA) and

covered with a glass petri dish no more than an hour before the panel.

171

Attribute references were prepared daily and placed into 2 or 4 oz. lidded plastic soufflé cups (Dart

Container Corporation, Mason, MI) and labeled with the reference identity. References were not

prepared more than 24 hours prior to evaluation. A complete list of attributes, definitions,

references, reference scores and preparation procedures can be found in Table 7.1. Product

information for references can be found in the Appendix L.

Subject Recruitment

All materials related to panelists recruitment and compensation were approved by the Institutional

Review Board (IRB) at the University of Illinois Urbana-Champaign (IRB Protocol Number: 16854),

Appendix D. The same panelists that were selected for the rum descriptive analysis panel in chapter

5 were retained for this descriptive analysis panel. Information regarding the recruitment and

screening process can be found there.

Descriptive Analysis Procedure

A hybrid of Qualitative Descriptive Analysis® (Stone, 1992) and the SpectrumTM method (Meilgaard

et al., 2007; Munoz & Civille, 1992) was used. Two rums were evaluated as a dilution series. At each

session, panelists received one dilution series consisting of three samples: straight rum, a 1:2 dilution

with 40%ABV ethanol, and a 1:2 dilution with water. On the first day of the panel, the panelists

were refreshed about the DA method to be used in the study. The first four sessions (1 hour each)

consisted of term and reference generation, followed by reference refinement. Panelists were

presented with a dilution series of samples labeled with random three digit codes and asked to

generate the attributes they perceived in the rum samples for aroma, aroma-by-mouth, mouthfeel,

taste and aftertaste modalities. Panelists were provided with the developed rum flavor wheel to aid in

term generation. Through group discussion, panelists identified the terms to be used, developed a

precise definition of each attribute and determined a corresponding reference.

After the terms and references were established, panelists then spent two sessions (1 hour each)

determining the reference intensities of each attribute. Panelists were asked to scale the references

based on a 15 point scale, where zero is no perception of a given attribute and 15 is the strongest

perception of that attribute in the rum samples. Panelists then spent six days practicing scoring the

rums, using the references as anchor points for the scale to aid in panel uniformity. On one of the

days, panelists conducted individual booth practice sessions in Bevier Hall on the University of

172

Illinois at Urbana-Champaign campus for one thirty minute session using the Compusense five

(Version 5.0: Guelph ON, Canada) data acquisition system. Panelists were routinely provided with

their scoring results from the previous day to help identify and correct for attributes they were rating

inconsistently with the rest of the group.

The panel concluded with two days of individual booth testing. Panelists attended two 30 minute

sessions per day, evaluating one dilution series of three samples at each session. Rum samples were

presented in black double old fashions glasses covered with a petri dish and labeled with random

three-digit codes. Sample presentation was randomized between all panelists. Samples were

evaluated under red lighting using the Compusense five software. Panelists were provided with their

reference tray when they arrived and encouraged to evaluate all references before proceeding into

the booth for testing. Panelists were free to leave the booth at any time to re-evaluate a reference.

A more detailed, day-by-day description of the DA panel can be found in Appendix P.

Statistical Analysis


SAS Institute Inc., Cary, NC, USA). Analysis of variance (ANOVA) was conducted on each of the

23 attributes evaluated during the DA panel to determine the presence of overall significant

differences (p<0.05) using the PROC GLM function for variations within the dilutions, panelists,

replications and their corresponding interactions: dilution-by-panelist (DxP), dilution-by-replication

(DxR), and replications-by-panelist (RxP). Each rum dilution series was analyzed separately. The

calculated probabilities were compared to significance levels α=0.05, 0.01 and 0.001. When

significant DxP interactions existed, adjusted F-ratios were calculated using Microsoft® Excel®

2016 (Version 16: Microsoft Corporation, Redmond, WA) by dividing the dilution mean square by

the DxP interaction mean square and calculating the new probability using the F.DIST function.

Fisher’s least significant difference (LSD) test was conducted on all attributes determined as

significant by ANOVA.

Principal component analysis (PCA) biplots were produced using SAS and Microsoft Excel to create

a visual representation of the data to allow further examination of the relationship of the rums to

individual attributes that characterized the samples. Pearson correlations were calculated using the

same SAS software, with significance determined at α=0.05, 0.01, and 0.001.

173


A descriptive analysis panel was performed to characterize the effect ethanol concentration has on

aroma perception of alcoholic beverages. A series of three samples was evaluated for each rum

consisting of the straight rum (directly from the bottle, 40%ABV), a 1:2 dilution with water (creating

a 20%AVB samples) to mimic how samples are routinely evaluated in industry, and a 1:2 dilution

with 40% ethanol to account for the flavor dilution effect but keep the alcohol concentration the

same. Two different rums were evaluated to assess if the effects were sample specific or applicable

to the larger body of distilled spirits. This descriptive analysis panel is the first sensory study to look

at the effects of ethanol concentration on flavor perception in distilled beverages.

The panelists generated 23 attributes to describe the two dilution series. The generated terms, term

definitions, selected references, reference scores and reference preparations can be found in Table

7.1. Reference scores are an average of individual panelist’s ratings. All panelists had previous

training evaluating the sensory properties of distilled beverages.

Analysis of variance (ANOVA) was conducted on the dilution series for each rum (DR and RA)

separately for all 23 attributes identified and rated by the panelists. The results are shown in Table

7.2 and Table 7.5. Overall, sample replications were not a significant source of error (p>0.05) for

either the DR12 dilution series (except silky mouthfeel and sweet taste) or the RA7 dilution series

(except toasted aroma, woody aroma, sweet taste and plastic aftertaste). This lack of variation shows

the panelists were able to rate the sample attributes across replications consistently. Significant

variation did exist (p<0.05) for all attributes for the DR12 dilution series (except alcohol aroma) and

RA7 dilution series. This type of variation is typical of descriptive analysis panels and is most likely a

result of panelists not using the entire scale or using different parts of the scale to rate the samples

(Lawless & Heymann, 1999b; Stone, Sidel, Oliver, Woolsey, & Singleton, 1974). R x P interactions

were not significant (p>0.05) for any attributes in the DR12 series nor for most attributes in the

RA7 series, except for warming mouthfeel, bitter taste, sweet taste, plastic aftertaste and maple

aroma by-mouth. The lack of interaction indicates that panelists were able to agree on the intensity

of the attributes in the samples across replications. There were no D x R effects in either series

(except slick mouthfeel in RA7 series) indicating that panelists rated the samples similarly across

replications.

174

Significant D x P interactions (p<0.05) did exist for DR12 series (except silky mouthfeel, sweet taste,

and vanilla aftertaste) and the RA7 series (except toasted aroma, silky mouthfeel, slick mouthfeel,

sweet taste, maple aroma-by-mouth and vanilla aroma by mouth). This interaction indicated that

panelists were not able to agree on the order of the intensity of the attributes across samples.

Adjusted F-vales were calculated for attributes that had significant D x R interaction to account for

the variation. The adjusted F-values are shown in Tables 7.2 and 7.5. After the adjusted F-values

were calculated, eighteen of the attributes in the DR12 series were determined to be significantly

different (p<0.05) including alcohol aroma, caramel aroma, maple aroma, vanilla aroma, roasted

aroma, woody aroma, astringent mouthfeel, silky mouthfeel, slick mouthfeel, warming mouthfeel,

bitter taste, bitter aftertaste, brown spice aftertaste, alcohol aroma-by-mouth, caramel aroma-by-

mouth, maple aroma-by-mouth, vanilla aroma-by-mouth, and woody aroma-by-mouth. In the RA7

dilution series sixteen attributes were determined to be significantly different (p<0.05) including

alcohol aroma, caramel aroma, vanilla aroma, dark fruit aroma, astringent mouthfeel, slick

mouthfeel, warming mouthfeel, bitter taste, bitter aftertaste, brown spice aftertaste, vanilla aftertaste,

plastic aftertaste, alcohol aroma-by-mouth, caramel aroma-by-mouth, maple aroma-by-mouth, and

vanilla aroma-by-mouth. Of the 23 attributes evaluated, all terms (except toasted aroma) were

statistically different for at least one of the rum series indicating that proper attributes were chosen

for evaluation.

Attribute Differences between Dilutions

The dilution of the rum samples, both with water and ethanol (40% ABV), caused significant

changes to the sensory profiles of the rums. The results indicated that rum samples diluted with

ethanol had the lowest intensities for all attributes (except silky mouthfeel in the DR12 series).

In the DR12 dilution series, the original undiluted rum was significantly different from DR40 for all

attributes, having a higher intensity for all attributes except silky mouthfeel (Table 7.3). Additionally,

the two dilutions, DR20 and DR 40, were also significantly different from each other in most

attributes, except caramel aroma, maple aroma, vanilla aroma, maple aroma-by-mouth and vanilla

aroma by mouth. DR20 had a higher intensity of all attributes except silky mouthfeel. DR and DR20

were not significantly different from each other except for caramel aroma, maple aroma, vanilla

aroma and brown spice aftertaste. Selected significant attribute correlations for the DR12 dilution

series samples (Table 7.4) include those between astringent mouthfeel and roasted aroma, warming

175

mouthfeel and alcohol aroma-by-mouth, and a significant negative correlation existed between silky

mouthfeel and roasted aroma, astringent mouthfeel, warming mouthfeel and alcohol aroma-by-

mouth.

In the RA7 dilution series, the original rum sample (RA) had the highest intensity of the three

samples for all attributes (Table 7.6). RA was significantly different from RA40 for all attributes.

RA40 had significantly lower attribute intensities compared to RA20 for all attributes except vanilla

aroma, vanilla aftertaste, and plastic aftertaste. RA and RA20 were significantly different from each

other for alcohol aroma, dark fruit aroma, astringent mouthfeel, caramel aroma-by-mouth, maple

aroma-by-mouth and vanilla aroma-by-mouth. Selected significant attribute correlations for the RA7

dilution series samples (Table 7.7) include those between bitter taste and warming mouthfeel and

alcohol aroma-by-mouth, slick mouthfeel with brown spice aftertaste, plastic aftertaste and alcohol

aroma-by-mouth, and astringent mouthfeel with slick mouthfeel, bitter, brown spice, and plastic

aftertaste, and alcohol and maple aroma-by-mouth.

Interestingly, rums diluted with water possessed nearly the same sensory profile as the original rums,

with slightly lower attribute intensities as shown in the constructed spider plots (Figures 7.1 and 7.3).

In contrast, the results indicate that dilution with ethanol profoundly changes the sensory profile of

the rum, especially RA7.

Principal component analysis was also conducted to reduce the complexity of the data and gain a

better visual representation of the data (Lawless & Heymann, 1999a). The covariance matrix was

chosen for sample evaluation since the samples were all scored by a trained panel (Jolliffe, 2014). For

the DR12 dilution series (Figure 7.2), the first factor (PC1) contained the majority of the variation

between samples (88.4%) and the second factor (PC2) contained the remaining sample variation

(11.6%). The high loading of factor one is also illustrated by the factor correlations (Appendix M).

PC1 contrasts samples high in brown sugar, caramel, maple, vanilla, coconut and chocolate aroma,

brown spice and caramel aftertaste, caramel, maple, vanilla and coconut aroma-by-mouth, with

samples high in alcohol, citrus, and phenolic aroma, warming mouthfeel and bitter taste. PC2 was

mainly defined by slick mouthfeel.

Focusing on the RA7 dilution series (Figure 7.4) the first factor contained the majority of the

variation as well (94.8%), with the second factor minimally loaded with the remaining variation

(5.2%). Factor loading can be found in Appendix N. PC1 contracted samples that were high in all

176

significant attributes with those that had lower intensities of those attributes. PC2 was defined by

samples differentiated in dark fruit aroma.

These results validate the initial hypothesis that the dilution with lower alcohol concentration would

have lower attribute intensity, particularly regarding aroma. Previous analytical studies on dynamic

systems showed that higher ethanol concentration had higher headspace concentrations of volatiles

(Taylor et al., 2010; Tsachaki et al., 2008, 2005). In the current study, the original rum sample at 40%

ABV had the highest intensity for all aroma attributes for both rum samples.

However, the results are not as significant as might have been expected, especially since both the

alcohol concentration and the congener concentrations were cut in half. No previous work

evaluating the effects of ethanol on flavor perception has considered the dilution effect that occurs

when a consumer dilutes the beverage. These studies have made replicate model solutions that are

identical except for their alcohol concentration (Boothroyd et al., 2012; Tsachaki et al., 2008, 2006,

2005; Tsachaki, Linforth, & Taylor, 2009). Even studies that evaluated wine model systems at

various ethanol concentrations spiked the solutions with the same concentration of volatiles after

the ethanol dilutions were prepared (Tsachaki et al., 2009). It is likely that the decrease in ethanol

concentration decreased the polarity of the system. Additionally, the evaporation effect and

subsequent stirring by the Marangoni effect and Rayleigh-Bénard convection (Marangoni, 1865;

Rayleigh, 1916) would still occur at 20% ABV, as it has been previously demonstrated in 12% ABV

systems (Taylor et al., 2010), and may account for why the aroma attribute intensities did not

decrease as much as was expected.

Interestingly, the 40% ABV dilutions were expected to be the closest to the original rums. However,

they had the lowest intensity for all attributes (except silky mouthfeel). Furthermore, the 40% ABV

dilutions were expected to have the highest perceived intensity of alcohol aroma, alcohol aroma-by-

mouth, and warming mouthfeel. While these results are opposite of what we expected, this is the

first sensory study to evaluate alcoholic beverages at such high alcohol concentrations. More sensory

studies are needed to confirm these results and gain a better understanding of how ethanol affects

sensory perceptions at higher alcohol levels.

In mouth sensory perceptions, including mouthfeel and taste, also differed as a result of dilution.

Regarding mouthfeel, the warming sensation was the same between the original rums and dilutions

with water but significantly decreased in the 40% ABV dilution. This is surprising since previous

177

research has demonstrated that increased ethanol concentration caused a higher rating of hotness or

burning mouthfeel sensation (Demiglio & Pickering, 2008; Jones et al., 2008; Nolden & Hayes,

2015). It was expected that the 40% ABV dilution would have had the highest warming sensation

followed by the original rum and then the dilution with water. It may be that ethanol, while one

factor contributing to the warming sensation of spirits, may not be the only chemical contributing to

that perception. These results are interesting as the ethanol used as the reference for warming

mouthfeel was the same ethanol used to dilute the samples.

Additionally, previous research demonstrated that increased ethanol concentration causes a decrease

in astringency (Demiglio & Pickering, 2008; Fontoin, Saucier, Teissedre, & Glories, 2008).

Interestingly, the 40% ABV dilution did have the lowest perceived astringency, except for the RA7

series, the water dilution was much less intense than the original rum. This difference could be

attributed to the fact that previous studies focused on wines. The high concentration of tannins

present in wines may alter the perception of astringency differently than distilled spirits when the

ethanol concentration changes.

Bitter taste was also shown to be significantly lower for the 40% ethanol dilutions in comparison to

the original rums and water dilutions. Previous studies have shown increases in ethanol

concentration shown to increase bitterness (Fontoin et al., 2008; Jones et al., 2008; Nolden & Hayes,

2015), which are contrary to our results. The other volatile and non-volatile components dissolved in

the rum matrix could account for these differences.

Sweetness was not shown to be significantly different between rum dilutions; however, previous

research has shown that sweetness perception increases with ethanol concentration in wines (Nurgel

& Pickering, 2006; Zamora, Goldner, & Galmarini, 2006) It is possible that the perceived sweetness

of rum is not large enough to enable panelists to perceive changes when the ethanol concentration

was changed.

This was the first study to evaluate the sensory effects of ethanol on distilled spirits. Our results

showed that the original rums and 1:2 dilutions with water were more similar to one another than

expected. The samples were only statistically different for several attributes in each series. These

results support the age old industry tradition of diluting distilled spirits to 20% or 23% ABV for

blending and evaluation, demonstrating that while the intensity of the attributes decreased in the

dilutions, the flavor profiles were very similar. The results for the 40% ABV samples was not

178

expected, and further research is needed to better understand how ethanol interacts with sensory

perceptions at high ethanol concentrations.

179


Table 7.1. List of final attributes, definitions, references, reference scores, and reference preparations determined for the descriptive analysis panel on

ethanol’s effect on the flavor perception of rums.

Modality Attribute Definition Reference Reference

Score Preparation

Aro

ma

alcohol aroma associated with ethanol 71 proof alcohol 11.5 10 oz. of (125 mL water + 75 ml 190 alcohol) in 2 oz.

cup

sugar

caramel aroma of caramelized sugar caramel sauce 12.5 2 g in 2 oz. cup

maple aroma of maple syrup maple extract 13.5 1 teaspoon in 500 mL volumetric flask, dilute to

volume, 10mL in 2 oz. cup

spices vanilla aroma of natural vanilla extract natural vanilla

extract 11


fruit dark fruit aroma associated with dried dark fruits prunes 14 0.4 g (~1/8) prune in a 2 oz. cup

woody

roasted aroma of medium roasted malted barley roasted barley 15 0.2 g in a 2 oz. cup

toasted a browned sweet aroma associated with

toasted marshmallow toasted

marshmallow 13

preheat boiler in oven, cut marshmallow in 1/8's, toast for 30 seconds and place 1/8 marshmallow in 4 oz.

cup

woody aroma of a wood barrel wood chips 10 0.5 g in 2 oz. cup

Mouthfeel

astringent a drying sensation in the mouth associated

with a high tannin wine over brewed

green tea 12

steep 1 tea bag in 300 mL of boiling water for 5 minutes, place ~15mL in a 2oz cup

silky an uninhibited flow of liquid over the

tongue, with a smooth feeling in the mouth almond milk 13 ~15 mL in a 2 oz. cup

slick a smooth tongue coating glycerin dilution 13.5 20 g of glycerin + 60 g water, ~10 g in a 2 oz. cup

warming the increase in temperature perception in

the mouth as a result of alcohol concentration

71 proof alcohol 14 10 oz. of (125 mL water + 75 ml 190 alcohol) in 2 oz.

cup

Taste

bitter taste associated with a caffeine solution caffeine solution 12.5 1 g caffeine in 500 mL of hot water, stir until

dissolved, ~15 mL in each cup

sweet taste associated with a sucrose solution sugar solution 11.5 4 g in 200 mL of water, stir, ~15 mL per cup

180

Table 7.1 (cont.) List of final attributes, definitions, references, reference scores, and reference preparations determined for the descriptive analysis panel

on ethanol’s effect on the flavor perception of rums

Modality Attribute Definition Reference Reference

Score Preparation

Aro

ma b

y M

ou

th

alcohol aroma-by-mouth associated with 40% or

greater alcohol 71 proof alcohol 14.5

10 oz. of (125 mL water + 75 ml 190 alcohol) in 2 oz. cup

sugar

caramel aroma-by-mouth of caramelized sugar caramel syrup 10.5 caramel solution, 10 g of caramel dissolved in 200 mL

of water, ~10 mL in a 2 oz. cup. Make daily

maple aroma-by-mouth of maple syrup maple extract 12 1 teaspoon in 500 mL volumetric flask, dilute to

volume, 10mL in 2 oz. cup

spices vanilla aroma-by-mouth of natural vanilla extract vanilla extract 12.5 1/4 teaspoon in 500 mL volumetric flask, dilute to

volume, 10 mL in 2 oz. cup woody aroma-by-mouth of a woody barrel wood chips 10.5 0.5 g in 2 oz. cup

Aft

ert

ast

e

bitter aftertaste associated with a caffeine

solution caffeine solution 12.5

1 g caffeine in 500 mL of hot water, stir until dissolved, ~15 mL in each cup

baking spices

brown spice

aftertaste associated with brown spices such as clove, and nutmeg

nutmeg 10 1g in 600 mL of water, stir 5 minutes, filter, 10 mL in

2 oz. cup

vanilla aftertaste associated with natural vanilla

extract natural vanilla

extract 10


plastic aftertaste associated with PVC plastic shower curtain 11 1 in x 1 in piece, place on tongue

181

Table 7.2 Analysis of variance (ANOVA) F-ratios for sensory attributes rated for dilutions of Diplomatico

Reserva 12-year rum+

+F-rations are shown as a source of variation. *, **, *** stand for significance at p<0.05, p<0.01, and p<0.001 respectively. †D x P, R x P, and D x R represent the interaction between dilution samples and panelists, replications and panelists, and dilution samples and replications, respectively.

Modality Attribute Dilution Panelist Rep DxP† DxR† RxP†

Adjusted Sample

F

Aroma

Alcohol 8.54** 2.16 0.12 0.81 1.33 0.12

Caramel 5.67* 4.18* 0.14 1.37 3.26 0.79

Maple 4.51* 11.48*** 0.21 1.48 2.91 1.23

Vanilla 5.64* 5.43** 0.5 1.19 1.51 0.7

Dark Fruit 1.26 8.29*** 0.07 0.82 0.38 0.53

Roasted 6.32* 8.15*** 0.07 0.71 0.46 1.19

Toasted 2.88 2.87* 0.03 0.59 0.52 0.97

Woody 8.15** 6.46** 3.02 1.49 2.77 1.62

Mouthfeel

Astringent 22.04*** 5.59** 0.64 0.82 0.28 1.53

Silky 22.41*** 6.62** 6.67* 3.94** 2.43 0.73 5.69*

Slick 14.6*** 3.81* 0.97 1.81 0.11 2.1

Warming 24.47*** 3.09* 0.27 0.62 0.12 0.37

Taste Bitter 20.56*** 5.9** 2.6 1.23 0.87 1.72

Sweet 5.31* 11.47*** 5.3* 4.67** 1.68 2.69 1.14

Aftertaste

Bitter 21.35*** 7.62*** 1.75 0.69 1.19 0.19

Brown Spice 11.1** 6.32** 0 0.9 0.54 1.05

Vanilla 2.3 3.47* 0.1 1.02 0.44 1.01 1.69

Plastic 7.22** 3.06* 1.56 4.28** 0.48 2.56

Aroma-

by-mouth

Alcohol 36.25*** 2.81* 0.06 0.93 0.36 0.35

Caramel 7.43** 3.09* 0.04 0.48 0.61 1.29

Maple 4.65* 2.88* 2.19 0.53 0.49 0.52

Vanilla 3.79* 5.47** 1.51 1.59 1.92 0.59

Woody 4.47* 3.29* 4.18 0.72 0.17 0.25

182

Table 7.3 Mean intensity rating for significant aroma, mouthfeel, taste, aftertaste and aroma-by-mouth

attributes of the Diplomatico Reserva 12 year rum dilution series†

Modality Attribute DR* DR20* DR40*

Aroma

Alcohol 9.44A 7.94A 5.19B

Caramel 10.25A 8.31B 7.56B

Maple 9.63A 7.81B 7.94B

Vanilla 10.56A 8.44B 8.50B

Roasted 5.00A 4.88A 3.19B

Woody 6.38A 5.75A 4.19B

Mouthfeel

Astringent 9.81A 9.31A 5.56B

Silky 6.63B 6.94B 10.44A

Slick 8.13A 7.00A 4.56B

Warming 9.19A 9.06A 3.81B

Taste Bitter 8.25A 8.50A 4.69B

Aftertaste Bitter 9.38A 9.71A 5.69B

Brown Spice 8.38A 7.00B 5.38C

Aroma-by-mouth

Alcohol 10.06A 9.50A 3.81B

Caramel 9.75A 8.75A 6.94B

Maple 8.75A 7.56A,B 6.44B

Vanilla 10.19A 9.06A,B 8.38B

Woody 6.63A 6.75A 4.75B

†Superscripts of the same letter within an attribute indicate no significant difference by Fisher’s Least Significant Difference (LSD) test at α=0.05. *“DR” is Diplomatico Reserva straight rum, “DR20” is 1:2 dilution of Diplomatico Reserva with water to achieve 20% ABV, “DR40” is 1:2 dilution of Diplomatico Reserva with 40% ethanol to achieve 40% ABV

183

Figure 7.1 Spider plot of mean significant attribute intensities the Diplomatico Reserva 12 year rum dilution series†* †“DR” is Diplomatico Reserva straight rum, “DR20” is 1:2 dilution of Diplomatico Reserva with water to achieve 20% ABV, “DR40” is 1:2 dilution of Diplomatico Reserva with 40% ethanol to achieve 40% ABV. * “A” is aroma, “ABM” is aroma-by-mouth,

“AT” is aftertaste, “MF” is mouthfeel, “Ta” is taste.

0.00

2.00

4.00

6.00

8.00

10.00

12.00

Brown Spice_AT

Caramel_A

Caramel_ABM

Maple_A

Maple_ABM

Vanilla_A

Vanilla_ABM

Slick_MF

Silky_MF

Astringent_MF

Warming_MF

Bitter_Ta

Bitter_AT

Alcohol_A

Alcohol_ABM

Roasted_A

Woody_A

Woody_ABM

DR DR20 DR40

184

Figure 7.2 Principal component analysis biplot of significant attributes present on principal component 1

(PC1) and 2 (PC2) by the correlation matrix of mean significant attribute intensity rating across all three

Diplomatico Reserva 12 year dilution samples

Alcohol_A

Caramel_A

Maple_AVanilla_A

Roasted_A

Woody_A

Astringent_MF

Silky_MF

Slick_MF

Warming_MFBitter_Ta

Bitter_AT

BrownSpice_AT

Alcohol_ABM

Caramel_ABM

Maple_ABM

Vanilla_ABM

Woody_ABM

DR

DR20

DR40

-1.5

-1

-0.5

0

0.5

1

-1.5 -1 -0.5 0 0.5 1 1.5

PC

2 (

11

.6%

)

PC1 (88.4%)

Variables (axes PC1 and PC2: 100%)

185

Table 7.4 Pearson correlation coefficients for significant attributes for DR12 dilution series samples

Attributes Alcohol

A Caramel

A Maple

A Vanilla

A Roasted

A Woody

A Astringent

MF Silky MF

Slick MF

Warming MF

Alcohol_A 1.000

Caramel_A 0.914 1.000

Maple_A 0.728 0.944 1.000

Vanilla_A 0.754 0.956 0.999* 1.000

Roasted_A 0.956 0.756 0.496 0.530 1.000

Woody_A 0.997* 0.882 0.676 0.705 0.975 1.000

Astringent_MF 0.970 0.786 0.537 0.570 0.999* 0.985 1.000

Silky_MF -0.960 -0.765 -0.508 -0.541 -1.000** -0.978 -0.999* 1.000

Slick_MF 0.999* 0.897 0.699 0.727 0.968 1.000* 0.979 -0.971 1.000

Warming_MF 0.945 0.730 0.462 0.497 0.999* 0.966 0.996* -0.999* 0.957 1.000

Bitter_Ta 0.915 0.673 0.390 0.426 0.993 0.942 0.986 -0.991 0.931 0.997

Bitter_AT 0.909 0.662 0.376 0.412 0.991 0.937 0.984 -0.989 0.925 0.995

BrownSpice_AT 0.993 0.956 0.806 0.828 0.914 0.981 0.932 -0.919 0.987 0.898

Alcohol_ABM 0.963 0.770 0.515 0.548 1.000* 0.980 1.000* -1.000** 0.973 0.998*

Caramel_ABM 1.000** 0.915 0.730 0.756 0.956 0.997 0.969 -0.960 0.999* 0.944

Maple_ABM 0.983 0.973 0.842 0.862 0.886 0.967 0.908 -0.893 0.975 0.868

Vanilla_ABM 0.952 0.994 0.903 0.919 0.821 0.928 0.848 -0.829 0.939 0.799

Woody_ABM 0.917 0.675 0.393 0.429 0.993 0.943 0.987 -0.992 0.932 0.997*

186

Table 7.4 (cont.) Pearson correlation coefficients for significant attributes for DR12 dilution series samples

Attributes Bitter

T Bitter

AT BrownSpice

AT Alcohol ABM

Caramel ABM

Maple ABM

Vanilla ABM

Woody ABM

Alcohol_A

Caramel_A

Maple_A

Vanilla_A

Roasted_A

Woody_A

Astringent_MF

Silky_MF

Slick_MF

Warming_MF

Bitter_Ta 1

Bitter_AT 1.000** 1

BrownSpice_AT 0.85969 0.85177 1

Alcohol_ABM 0.99025 0.988 0.92246 1

Caramel_ABM 0.91416 0.90784 0.99296 0.96171 1

Maple_ABM 0.82546 0.81672 0.998* 0.89604 0.98341 1

Vanilla_ABM 0.7486 0.73836 0.98224 0.83365 0.9531 0.99219 1

Woody_ABM 1.000** 1.000* 0.86116 0.99065 0.91532 0.82708 0.7505 1


187

Table 7.5 Analysis of variance (ANOVA) F-ratios for sensory attributes rated for dilutions of Ron Abuelo 7 year

rum+

+F-rations are shown as a source of variation. *, **, *** stand for significance at p<0.05, p<0.01, and p<0.001 respectively. †D x P, R x P, and D x R represent the interaction between dilution samples and panelists, replications and panelists, and dilution samples and replications, respectively.

Modality Attribute Dilution Panelist Rep DxP† DxR† RxP†

Adjusted

Sample

F

Aroma

Alcohol 40.85*** 8.71*** 0.23 2.03 1.22 2.37

Caramel 5.94* 11.54*** 0.81 1.20 0.15 1.08

Maple 3.66 8.94*** 1.06 0.94 0.67 2.30

Vanilla 5.36* 8.96*** 2.64 1.37 1.48 1.10

Dark Fruit 3.98* 12.04*** 0.16 1.48 0.66 0.80

Roasted 2.38 4.15* 0.52 1.42 0.51 1.68

Toasted 0.72 15.38*** 5.56*** 4.18** 1.17 1.58 0.17

Woody 0.18 9.18*** 4.61*** 1.94 0.08 1.89

Mouthfeel

Astringent 15.17*** 15.17*** 0.24 1.56 0.47 2.18

Silky 10.09*** 6.85*** 1.04 3.19*** 2.01 1.18 3.17

Slick 20.36*** 7.36*** 0.51 2.96* 4.44* 2.68 6.87**

Warming 59.21*** 11.57*** 0.06 1.17 1.20 3.02***

Taste Bitter 49.87*** 13.75*** 0.08 0.78 0.84 6.62***

Sweet 3.60 7.58*** 6.62* 3.26* 1.00 4.73** 1.11

Aftertaste

Bitter 29.14*** 11.98*** 2.28 1.11 0.77 2.30

Brown Spice 11.79** 5.91** 2.51 1.53 0.01 1.25

Vanilla 5.62* 11.8*** 0.34 1.80 0.26 1.28

Plastic 4.34* 12.72*** 6.42* 2.23 3.26 3.38*

Aroma-

by-mouth

Alcohol 47.80*** 7.54*** 2.29 0.57 1.37 0.66

Caramel 16.90*** 9.47*** 0.01 2.08 1.50 0.37

Maple 26.65*** 17.89*** 1.18 3.41* 0.68 3.72* 7.81**

Vanilla 23.49*** 33.16*** 1.20 2.58* 0.93 1.57 9.10**

Woody 1.56 13.48*** 0.38 1.77 0.43 0.78

188

Table 7.6 Mean intensity rating for significant aroma, mouthfeel, taste, aftertaste and aroma-by-mouth

attributes of the Ron Abuelo 7 year rum dilution series†

Modality Attribute RA* RA20 RA40

Aroma Alcohol 9.63A 7.94B 4.88C

Caramel 8.13A 7.31A 5.81B

Vanilla 8.44A 7.50A,B 6.19B

Dark Fruit 7.13A 5.13B 5.50B

Mouthfeel Astringent 10.25A 8.81B 4.75C

Slick 6.63A 5.75A 3.13B

Warming 9.63A 9.06A 3.63B

Taste Bitter 9.06A 8.38A 4.13B

Aftertaste Bitter 10.25A 9.06A 6.00B

Brown Spice 7.56A 7.13A 5.63B

Vanilla 7.06A 6.25A,B 5.31B

Plastic 4.94A 4.50A,B 3.44B

Aroma-by-mouth Alcohol 10.81A 9.38A 3.94B

Caramel 8.44A 7.06B 5.06C

Maple 8.06A 7.19B 5.38C

Vanilla 8.63A 7.19B 5.75C

†Superscripts of the same letter within an attribute indicate no significant difference by Fisher’s Least Significant Difference (LSD) test at α=0.05. *“RA” is Ron Abuelo 7 year straight rum, “RA20” is 1:2 dilution of Ron Abuelo 7 year with water to achieve 20% ABV, “RA40” is 1:2 dilution of Ron Abuelo 7 year with 40% ethanol to achieve 40% ABV.

189

Figure 7.3 Spider plot of mean significant attribute intensities the Ron Abuelo 7 year rum dilution series†*

†“RA” is Ron Abuelo 7 year straight rum, “RA20” is 1:2 dilution of Ron Abuelo 7 year with water to achieve 20% ABV, “RA40” is

1:2 dilution of Ron Abuelo 7 year with 40% ethanol to achieve 40% ABV. * “A” is aroma, “ABM” is aroma-by-mouth, “AT” is

aftertaste, “MF” is mouthfeel, “Ta” is taste.

0.00

2.00

4.00

6.00

8.00

10.00

12.00

Brown Spice_AT

Caramel_A

Caramel_ABM

Maple_ABM

Vanilla_A

Vanilla_ABM

Vanilla_AT

Plastic_AT

Slick_MF

Warming_MF

Alcohol_ABM

Alcohol_A

Astringent_MF

Bitter_AT

Bitter_Ta

Dark Fruit_A

RA RA20 RA40

190

Figure 7.4 Principal component analysis biplot of significant attributes present on principal component 1

(PC1) and 2 (PC2) by the correlation matrix of mean significant attribute intensity rating across all three Ron

Abuelo 7 year dilution samples

Alcohol_ACaramel_A Vanilla_A

DarkFruit_A

Astringent_MFSlick_MF

Warming_MFBitter_Ta

Bitter_ATBrownSpice_AT

Vanilla_AT

Plastic_AT

Alcohol_ABM

Caramel_ABM

Maple_ABM

Vanilla_ABM

RA

RA20

RA40

-1.5

-1

-0.5

0

0.5

1

-1.5 -1 -0.5 0 0.5 1 1.5

PC

2 (

5.2

%)

PC1 (94.8%)

Variables (axes PC1 and PC2: 100%)

191

Table 7.7 Pearson correlation coefficients for significant attributes for RA7 dilution series samples

Attributes Alcohol

A Caramel

A Vanilla

A DarkFruit

A Astringent

MF Slick MF

Warming MF

Bitter T

Bitter AT

BrownSpice AT

Alcohol_A 1.000

Caramel_A 1.000** 1.000

Vanilla_A 0.998* 0.997* 1.000

DarkFruit_A 0.650 0.648 0.702 1.000

Astringent_MF 0.995 0.995 0.985 0.568 1.000

Slick_MF 0.993 0.994 0.983 0.559 1.000** 1.000

Warming_MF 0.963 0.964 0.942 0.421 0.986 0.988 1.000

Bitter_Ta 0.973 0.974 0.955 0.459 0.992 0.993 0.999* 1.000

Bitter_AT 0.996 0.997 0.988 0.584 1.000* 1.000 0.982 0.989 1.000

BrownSpice_AT 0.990 0.990 0.977 0.533 0.999* 1.000* 0.992 0.996 0.998* 1.000

Vanilla_AT 0.993 0.992 0.999* 0.738 0.975 0.972 0.923 0.938 0.979 0.965

Plastic_AT 0.998* 0.998* 0.990 0.596 0.999* 0.999* 0.979 0.987 1.000** 0.997*

Alcohol_ABM 0.987 0.988 0.974 0.520 0.998* 0.999* 0.994 0.997* 0.997* 1.000**

Caramel_ABM 0.998* 0.998* 1.000** 0.694 0.987 0.985 0.945 0.958 0.990 0.979

Maple_ABM 0.999* 0.999* 0.994 0.623 0.998* 0.997 0.972 0.981 0.999* 0.994

Vanilla_ABM 0.986 0.986 0.996 0.766 0.964 0.961 0.906 0.923 0.969 0.952

192

Table 7.7 (cont.) Pearson correlation coefficients for significant attributes for RA7 dilution series samples

Attributes Vanilla

AT Plastic

AT Alcohol ABM

Caramel ABM

Maple ABM

Vanilla ABM

Alcohol_A

Caramel_A

Vanilla_A

DarkFruit_A

Astringent_MF

Slick_MF

Warming_MF

Bitter_Ta

Bitter_AT

BrownSpice_AT

Vanilla_AT 1.000

Plastic_AT 0.982 1.000

Alcohol_ABM 0.960 0.996 1.000

Caramel_ABM 0.998* 0.992 0.976 1.000

Maple_ABM 0.988 0.999* 0.992 0.996 1.000

Vanilla_ABM 0.999* 0.973 0.948 0.994 0.980 1.000

*, **, *** stand for significance at p<0.05, p<0.01, and p<0.001 respectively

193

7.6 References

Athès, V., Pena-Lillo, M., Bernard, C., Perez-Correa, R., & Souchon, I. (2004). Comparison of

Experimental Methods for Measuring Infinite Dilution Volatilities of Aroma Compounds in

Water/Ethanol Mixtures. Journal of Agricultural and Food Chemistry, 52, 2021–2027.







Cipiciani, A., Onori, G., & Savelli, G. (1988). Structural Properties of Water-Ethanol Mixtures: A

Correlation with the Formation of Micellar Aggregates. Chemical Physics Letters, 143(5), 505–509.

Conner, J. M., Birkmyre, L., Paterson, A., & Piggott, J. R. (1998). Headspace Concentrations of

Ethyl Esters at Different Alcoholic Strengths. Journal of the Science of Food and Agriculture, 77,

121–126.

D’Angelo, M., Onori, G., & Santucci, A. (1994a). Self-Association Behaviour of Alcohols in Diluted

Aqueous Solutions. Il Nuovo Cimento D, 16(9), 1499–1514.

http://doi.org/10.1007/BF02462035

D’Angelo, M., Onori, G., & Santucci, A. (1994b). Self-Association of Monohydric Alcohols in

Water: Compressibility and Infrared Absorption Measurements. The Journal of Chemical Physics,

100(4), 3107–3113. http://doi.org/10.1063/1.466452

Demiglio, P., & Pickering, G. J. (2008). The influence of ethanol and pH on the taste and mouthfeel

sensations elicited by red wine. Journal of Food, Agriculture and Environment, 6(3–4), 143–150.

Escalona-Buendia, H., Piggott, J. R., Conner, J. M., & Paterson, A. (1998). Effect of ethanol strength

on the release of higher alcohols and aldehydes in model solutions. Developments in Food Science,

40(C), 615–620.

194

Escudero, A., Campo, E., Fariña, L., Cacho, J., & Ferreira, V. (2007). Analytical Characterization of

the Aroma of Five Premium Red Wines. Insights into the Role of Odor Families and the

Concept of Fruitiness of Wines. Journal of Agricultural and Food Chemistry, 55, 4501–4510.

http://doi.org/10.1021/jf0636418

Fontoin, H., Saucier, C., Teissedre, P. L., & Glories, Y. (2008). Effect of pH, ethanol and acidity on

astringency and bitterness of grape seed tannin oligomers in model wine solution. Food Quality

and Preference, 19(3), 286–291. http://doi.org/10.1016/j.foodqual.2007.08.004

Franks, F., & Ives, D. J. G. (1966). The Structural Properties of Alcohol-Water Mixtures. Quarterly

Reviews, 1–44.

Goldner, M. C., Zamora, M. C., Di Leo Lira, P., Gianninoto, H., & Bandoni, A. (2009). Effect of

Ethanol Level in the Perception of Aroma Attributes and the Detection of Volatile

Compounds in Red Wine. Journal of Sensory Studies, 24(2), 243–257.

http://doi.org/10.1111/j.1745-459X.2009.00208.x

Guth, H. (1998). Comparison of Different White Wine Varieties in Odor Profiles by Instrumental

Analysis and Sensory Studies. ACS Symposium Series, 714, 39–52. Retrieved from

http://www.scopus.com/inward/record.url?eid=2-s2.0-

0347804788&partnerID=40&md5=d77c8f8f9d686ec570e3e57b7c623e13

Jolliffe, I. (2014). Principal component analysis. In Wiley StatsRef: Statistics Reference Online (pp. 1–5).

Jones, P. R., Gawel, R., Francis, I. L., & Waters, E. J. (2008). The influence of interactions between

major white wine components on the aroma, flavour and texture of model white wine. Food

Quality and Preference, 19(6), 596–607. http://doi.org/10.1016/j.foodqual.2008.03.005

King, E. S., Dunn, R. L., & Heymann, H. (2013). The influence of alcohol on the sensory perception

of red wines. Food Quality and Preference, 28(1), 235–243.

http://doi.org/10.1016/j.foodqual.2012.08.013

Lawless, H. T., & Heymann, H. (1999a). Data Relationships and Multivariate Applications. In Sensory

Evaluation of Food: Principles and Practices (pp. 585–601). New York: SPRINGER

SCIENCE+BUSINESS MEDIA, LLC.

195

Lawless, H. T., & Heymann, H. (1999b). Descriptive analysis. In Sensory Evaluation of Food: Principles

and Practices (pp. 227–257). New York: SPRINGER SCIENCE+BUSINESS MEDIA, LLC.

http://doi.org/10.1007/978-1-4419-6488-5

Le Berre, E., Atanasova, B., Langlois, D., Etiévant, P., & Thomas-Danguin, T. (2007). Impact of

ethanol on the perception of wine odorant mixtures. Food Quality and Preference, 18(6), 901–908.

http://doi.org/10.1016/j.foodqual.2007.02.004

Marangoni, C. (1865). On the expansion of a drop of liquid floating on the surface of another liquid. University

of Pavia, Pavia, Italy.

Meilgaard, M., Civille, G. V., & Carr, B. T. (2007). The Spectrum Descriptive Analysis Method. In

Sensory Evaluation Techniques (Fourth Edi, pp. 189–254). Boca Raton, FL: CRC Press/Taylor &

Francis.

Munoz, A. M., & Civille, G. (1992). The Spectrum Descriptive Analysis Method. In R. C. Hootman

(Ed.), Manual on Descriptive Analysis Testing for Sensory Evaluation (pp. 22–34). Baltimore, MD:

American Society for Testing and Materials.

Nolden, A. A., & Hayes, J. E. (2015). Perceptual Qualities of Ethanol Depend on Concentration,

and Variation in These Percepts Associates with Drinking Frequency. Chemosensory Perception,

8(3), 149–157. http://doi.org/10.1007/s12078-015-9196-5

Nurgel, C., & Pickering, G. (2006). Modeling of sweet, bitter and irritant sensations and their

interactions elicited by model ice wines. Journal of Sensory Studies, 21, 505–519.

Onori, G., & Santucci, A. (1996). Dynamical and structural properties of water/alcohol mixtures.

Journal of Molecular Liquids, 69, 161–181.

Parke, S. A., & Birch, G. G. (1999). Solution Properties of Ethanol in Water. Food Chemistry, 67, 241–

246.

Petrillo, C., Onori, G., & Sacchetti, F. (1989). Hydration structure of ethanol water solution at low

alcohol concentration. Molecular Physics, 67(3), 697–705.

http://doi.org/10.1080/00268978900101381

Rayleigh, Lord. (1916). LIX. On convection currents in a horizontal layer of fluid, when the higher

196

temperature is on the under side. Philosophical Magazine, 32(192), 529–546.

http://doi.org/10.1080/14786441608635602

Smith, G. D., & Roskrow, D. (2012). Whiskeyopus. New York: DK Publishing.

Stone, H. (1992). Quantitative Descriptive Analysis. In R. C. Hootman (Ed.), Manual on Descriptive

Analysis Testing for Sensory Evaluation (pp. 15–21). Baltimore, MD: American Society for Testing

and Materials.

Stone, H., & Sidel, J. L. (1993). Descriptive Analysis. In S. L. Taylor (Ed.), Sensory Evaluation Practices

(2nd ed., pp. 202–242). San Diego: Academic Press Inc.

Stone, H., Sidel, J. L., Oliver, S., Woolsey, A., & Singleton, R. (1974). Sensory Evaluation by

Quantitative Descriptive Analysis. Food Technology, 8(1), 24–32.

http://doi.org/10.1002/9780470385036.ch1f



2010-1036.ch012



4501(06)80104-4








Tsachaki, M., Linforth, R. S. T., & Taylor, A. J. (2009). Aroma Release from Wines under Dynamic

Conditions. Journal of Agricultural and Food Chemistry, 57(15), 6976–6981.

http://doi.org/10.1021/jf901174y

197

Wakisaka, A., Komatsu, S., & Usui, Y. (2001). Solute-solvent and solvent-solvent interactions

evaluated through clusters isolated from solutions: Preferential solvation in water-alcohol

mixtures. Journal of Molecular Liquids, 90, 175–184. http://doi.org/10.1016/S0167-

7322(01)00120-9

Williams, A. A. (1972). Flavour Effects of Ethanol in Alcoholic Beverages. The Flavour Industry, 3(12),

604–607.

Zamora, M. C., Goldner, M. C., & Galmarini, M. V. (2006). Sourness - Sweetness Interactions in

Different Media: White Wine, Ethanol and Water. Journal of Sensory Studies, 21, 601–611.

http://doi.org/10.1111/j.1745-459X.2006.00085.x

198

Chapter 8: Summary, Conclusions, and Future Research

Rum is one of the most diverse distilled spirits on the market. The breadth of variation within the

rum category makes it difficult to ascertain the key odor-active compounds that define rum flavor.

The goal of this study was to gain a better understanding of rum flavor through analytical and

sensory evaluation of mixing and premium aged rums. The central hypothesis of this study was that

premium aged rums, while produced using a variety of methods, still have a defining set of aroma

characteristics caused by a unique combination of flavor compounds that set these rums apart from

those of lower quality, or so-called mixing rums, and that ethanol concentration plays an important

role in the perception of overall rum flavor.

Identification of the odor-active compounds in the nine rums by gas chromatography-olfactometry

(GCO) and GC-mass spectrometry (GC-MS) analysis yielded 59 odor-active regions containing 64

odor-active compounds. Aroma extract dilution analysis (AEDA) provided a ranking of the potency

of odorants. Acetal (melon), 2-/3-methyl-1-butanol (chocolate), β-damascenone (applesauce), 2-

phenethyl alcohol (roses), cis-whiskey lactone/4-methylguaiacol (sweet, coconut-like), eugenol (spicy,

clove), sotolon (curry, maple-like), syringol (smoky, spicy), (E)-isoeugenol (floral, cloves), vanillin

(vanilla, sweet-like), ethyl vanillate (vanilla, sweet aromatic), and syringaldehyde (vanilla) were the

most potent odorants across all rums. Comparison of mixing and premium rums indicated that they

contained many of the same compounds, but generally present at lower potency in mixing rums.

Premium rums contained additions compounds that were not detected in the mixing rums. Fourteen

unknown odor-active regions were identified by GCO. Future research should focus on

identification of these unknowns, particularly unknown 59 (Wax RI-2951, campfire) as it was

identified in all nine rums and may contribute to the smoky and woody perception of rums.

Thirty-four of the compounds identified by GCO and AEDA were quantitated by stable isotope

dilution analysis. Quantitation results revealed the same compounds to be present in all rums

samples with the exception of 4-ethylguaiacol and eugenol in BW and ethyl vanillin which was only

detected in DR12 and DX. The mixing rums, BW and BG, and DX were found to have the lowest

concentrations of all compounds quantitated in the rums. Concentrations were converted to odor

activity values (OAVs) to gain a better understanding of the importance of the compounds to the

199

overall aroma of the rum. Twenty-six compounds were found to have OAVs >1 in at least one rum.

Fifteen compounds had OAVs >1 in all nine samples including 2-methylpropanal, acetal, 3-

methylbutanal, 2-methylbutanal, ethyl 2-methylpropanoate, ethyl butanoate, ethyl 2-

methylbutanoate, ethyl 3-methylbutanoate, 3-methyl-1-butanol, 2-methyl-1-butanol, ethyl hexanoate,

β-damascenone, guaiacol, cis-whiskey lactone, and vanillin. Several low potency odorants identified

by GCO were not quantitated due to time limitations. Quantitation of these compounds should be

the focus of future studies, particularly 1-octen-3-one, ethyl cyclohexanecarboxylate, methional, (Z)-

2-nonenal, 3-methylbutyric acid, 4-propylguaiacol, and sotolon.

A run flavor lexicon wax created to characterize the sensory differences among rum products. Web-

based materials were used to gather the terms for the lexicon to minimize the time and cost and

allow for the inclusion of a greater number of rums. This was the first lexicon developed using this

methodology. The final lexicon consisted of 147 terms sorted into 22 categories

Descriptive sensory analysis was then conducted to verify the rum flavor lexicon and quantitate the

sensory differences between nine rums previously evaluated by analytical measures. Thirty-three of

the thirty-eight terms used to evaluate the rums were found on the flavor wheel, validating that the

lexicon contained terms relevant to the sensory evaluation of rums. The additional five terms

generated during the panel demonstrate that the lexicon will need to be updated over time, which is

typical of other lexicons that have been developed for beer, wine, and coffee.

Twenty-three sensory attributes were found to be significantly different among the rums. Two rums,

DX and DR12, were found to be significantly different from the other seven rums, having higher

intensity ratings for brown sugar, caramel, vanilla and chocolate aroma, caramel, maple and vanilla

aroma-by-mouth, and caramel aftertaste. Sensory profiles of the other seven rums were very similar

to each other. Lack of differences observed between the other seven rums is likely a result of

panelists not using the entire scale, improper scoring of references, as well as sensory fatigue.

Additionally, several terms characterized different but similar attributes in rums such as caramel,

vanilla, and maple aroma that followed similar intensity ratings among rum samples. Future studies

on rum should focus on the evaluation of fewer attributes that characterize the largest differences

among rums. Terms that evaluate similar attributes should be reduced, allowing panelists to better

focus on the other sensory changes among rums. Additionally, panelists may be able to better

200

identify the difference among the rums if they were evaluated at 20% ABV rather than 40% ABV to

reduce ethanol pungency and sensory fatigue.

Descriptive sensory evaluation was also conducted to gain insight into the effect of ethanol on flavor

perception. Two rums, RA7 and DR12, were evaluated at three different dilution levels: straight

rum, 1:2 dilution with water, and a 1:2 dilution with 40%ABV. Dilutions of rum with water, while

hypothesized to alter the flavor profile of rum, yielded similar profiles to straight rum, except with

slightly lower attribute intensity ratings. However, dilution with 40% ethanol did significantly change

the profile of rum and also had the lowest intensity rating in the dilution series for most attributes.

This study validates the typical practice in the whiskey and rum industry of diluting samples to 20-

23% for blending and evaluation. Further sensory studies are needed to further explore the effects

of ethanol concentration on sensory perceptions. This study only focused on two levels of ethanol,

20% ABV and 40%ABV, where the concentrations of the aroma compounds were cut in half in

both dilutions relative to the straight rum. Creation of an accurate rum model system would allow

researchers to account for sensory changes caused by dilution of both ethanol and the aroma

constituents, as models could also be constructed that where two samples had the same

concentration of compounds and only differed in ethanol concentration rather than evaluation the

dilution of an actual rum as in the present study. Corresponding analytical studies following the

same samples should be included as well, primarily focusing on the release of compounds into the

headspace under dynamic conditions. Additional ethanol concentrations should be evaluated in the

future as well.

Finally, chemometric analysis was conducted to correlate the sensory and analytical data using

principal component analysis. Correlations between sensory evaluations, quantitation and OAV data

explained the most variation between rums, accounting for 68.6 % and 65.5% respectively. Results

indicate the changes in vanilla, caramel, maple and chocolate aromas are driven by vanillin and ethyl

vanillin. Additionally, roasted aroma is defined by an absence of compounds rather than increases in

concentration of any particular compounds. Overall, the correlations between sensory and

quantitative data did not identify as many correlations as originally thought. The lack of significant

correlations may be due to the minimal aroma differences perceived among the seven rums not high

in the sweet aroma sensory attributes. A more in-depth sensory analysis focusing on the difference

among rums other than sweet aroma attributes may aid in the identification of more analytical and

sensory relationships. Additionally, only nine rums were evaluated which is much lower than the

201

total number of variables used for the correlation. The inclusion of a greater variety and number of

rums may provide deeper insights.

In conclusion, the main differences between mixing and premium rums is the concentrations of

compounds, with mixing rums having lower concentrations of all compounds. This is reflected in

the similar aroma profiles obtained for the mixing and premium rums through descriptive analysis.

Rums that contained added ethyl vanillin, DR12 and DX, had higher perceptions of sweet-like

aroma and aroma-by-mouth attributes. Differences in concentration and ratios of compounds

relative to each other seem to be the driving forces behind the difference in flavor perception among

rums.

Future research is needed to characterize the compounds responsible for the distinct “rummy”

perception of rums. Rum contains many of the same compounds as other distilled spirits such as

whiskey, and a more in-depth analysis of rums high in “rummy” perception is needed. The creation

of an accurate rum model system would aid in the evaluation of both the effects of ethanol on flavor

perception and the importance of the identified compounds to overall rum flavor, specifically what

compounds characterize “rummy” flavor.

Finally, we are still only analyzed a relatively small subset of rums. The increased number of samples

compared to previous studies has allowed greater insight into the compounds that are significant

aroma contributors across all rums. However, compounds that may be present only in the lower

dilutions in these samples may be more important to other brands of rum. A broader study of all

types of rums is needed. A preliminary sensory evaluation and sorting exercise of a larger variety of

rum samples using techniques such as Napping to identify similar types of rums could be employed.

Selection of one rum from each category could then be subject to sensory and analytical evaluation

to provide greater insight into the differences among rums across the class.

202

Appendix A: Standard curves

Response Factor of acetaldehyde

Standard: acetaldehyde CAS: 75-07-0 Mfg/Reference: Sigma-Aldrich, St. Louis, MO No.: 627 % Purity (by GC-FID): 99.9% Standard Curve

acetaldehyde (density g/mL) 0.788

retention time (min) 2.12

Concentration

ug/mL Area

2uL 1:100 dilution 40ABV 15.76 14.3

5uL 1:100 dilution 40ABV 39.4 25.2

10uL 1:100 dilution 40ABV 78.8 57.3

2uL 1:10 dilution 40ABV 157.6 102.1

5uL 1:10 dilution 40ABV 394 313.6

slope Rf

0.7983 1.253

y = 0.7983x - 6.9631R² = 0.9935

0

50

100

150

200

250

300

350

0 100 200 300 400 500

Are

a C

ou

nt

Concentration (ug/mL)

203

Response Factor of d2-2-methylpropanal

Isotope Unlabeled Standard: d2-2-methylpropanal 2-methylpropanal CAS: N/A 78-84-2 Mfg/Reference: synthesized Aldrich, Milwaukee, WI No.: ISO-159 858 % Purity (by GC-FID): N/A 99.3% Standard Curve

selected ion mass ratio

unlabeled 72

labeled 74

area ratio

10.8 358925 19428 18.5

7.2 242780 19324 12.6

3.6 127238 19701 6.46

1.8 138396 45065 3.07

1.2 124167 60586 2.05

slope Rf

1.7287 0.579

y = 1.7287xR² = 0.9995

0

5

10

15

20

0 5 10 15

Are

a ra

tio

(io

n 7

2/i

on

74

)

Mass ratio (std/iso)

204

Response Factor of d10-acetal

Isotope Unlabeled Standard: d10-acetal acetal CAS: N/A 105-57-7 Mfg/Reference: synthesized Acros Organics, NJ No.: N/A 9 % Purity (by GC-FID): N/A 98.88% Standard Curve


unlabeled 103

labeled 113

area ratio

0.122 73612 412802 0.178

0.305 147584 358468 0.412

0.609 115742 108059 1.071

3.047 1450470 351010 4.132

6.095 3759819 502441 7.483

slope Rf

1.2207 0.518

y = 1.2207x + 0.1703R² = 0.9965

0

1

2

3

4

5

6

7

8

0 1 2 3 4 5 6 7

Are

a ra

tio

(io

n 1

03

/io

n 1

13

)


205

Response Factor of d2-3-methylbutanal

Isotope Unlabeled Standard: d2-3-methylbutanal 3-methylbutanal CAS: N/A 590-86-3 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: N/A 859 % Purity (by GC-FID): N/A 91.4% Standard Curve


unlabeled 71

labeled 73

area ratio

4.920 8014 80562 10.100

3.280 7920 53170 6.710

1.640 10195 34214 3.360

0.820 23412 41725 1.780

0.550 30803 33474 1.090

slope Rf

2.0479 0.488

y = 2.0479x + 0.0139R² = 0.9998

0

2

4

6

8

10

12

0 2 4 6

Are

a ra

tio

(io

n 7

1/i

on

73

)


206

Response Factor of d2-2-methylbutanal

Isotope Unlabeled Standard: d2-2-methylbutanal 2-methylbutanal CAS: N/A 96-17-3 Mfg/Reference: synthesized Bedoukian, Danbury, CT No.; Catalog#; Batch#/Lot#: N/A 1163 % Purity (by GC-FID): N/A 99.1% Standard Curve


unlabeled 86

labeled 88

area ratio

4.78 34323 3811 9.01

3.19 21287 3481 6.12

1.59 15324 4190 3.66

0.8 16128 9473 1.7

0.53 13548 12181 1.11

slope Rf

1.9294 0.518

y = 1.9294xR² = 0.9919

0

2

4

6

8

10

0 2 4 6

Are

a ra

tio

(io

n 8

6/i

on

88

)


207

Response Factor of d5-ethyl propanoate

Isotope Unlabeled Standard: d5-ethyl propanoate ethyl propanoate CAS: N/A 105-37-3 Mfg/Reference: CDN Aldrich, Milwaukee, WI No.; Catalog#; Batch#/Lot#: N/A 293 % Purity (by GC-FID): N/A 99.5% Standard Curve


unlabeled 57

labeled 62

area ratio

5.963 6114367 1126476 5.428

2.385 2054324 858676 2.392

1.193 1138318 786559 1.447

0.596 580622 726698 0.799

0.239 208686 907427 0.230

slope Rf

0.880 1.136

y = 0.88x + 0.2333R² = 0.9952

0

1

2

3

4

5

6

0 2 4 6 8

Are

a ra

tio

(io

n 5

7/i

on

62

)


208

Response Factor of d5-ethyl 2-methylpropanote

Isotope Unlabeled Standard: d5-ethyl 2-methylpropanoate ethyl 2-methylpropanoate CAS: N/A 97-62-1 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: ISO-81 341 % Purity (by GC-FID): N/A 98.8% Standard Curve


unlabeled 116

labeled 121

area ratio

4.896 103646 25720 4.030

1.959 28906 17989 1.607

0.979 20050 24695 0.812

0.490 11352 26637 0.426

0.196 4145 22665 0.183

slope Rf

0.8185 1.220

y = 0.8185x + 0.017R² = 1

0

1

2

3

4

5

0 1 2 3 4 5 6Are

a ra

tio

(io

n 1

16

/io

n 1

21

)


209

Response Factor of d7-ethyl butyrate

Isotope Unlabeled Standard: d7-ethyl butyrate ethyl butyrate CAS: N/A 105-54-4 Mfg/Reference: CDN Sigma-Aldrich, St. Louis, MO No.: ISO-17 283 % Purity (by GC-FID): N/A 92.7% Standard Curve


unlabeled 71

labeled 78

area ratio

2.263 1634048 401807 4.067

0.905 532332 344723 1.544

0.453 297583 407924 0.730

0.226 159134 411304 0.387

0.091 55615 363203 0.153

slope Rf

1.8086 0.553

y = 1.8086x - 0.0483R² = 0.9994

0

1

2

3

4

5

0.0 0.5 1.0 1.5 2.0 2.5Are

a ra

tio

(io

n 7

1/i

on

78

)


210

Response Factor of d5-ethyl 2-methylbutanoate

Isotope Unlabeled Standard: d2-ethyl 2-methylbutanoate ethyl 2-methylbutanoate CAS: N/A 7452-79-1 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: N/A 623 % Purity (by GC-FID): N/A N/A Standard Curve


unlabeled 102

labeled 107

area ratio

1.688 8111874 4700583 1.726

1.125 5125493 4461740 1.149

0.563 2445898 4306635 0.568

0.281 3052775 10920483 0.280

0.188 2946222 15657814 0.188

slope Rf

1.0204 0.980

y = 1.0204xR² = 0.9999

0.0

0.5

1.0

1.5

2.0

0.0 0.5 1.0 1.5 2.0

Are

a ra

tio

(io

n 1

02

/io

n1

07

)


211

Response Factor of d5-ethyl 3-methylbutanoate

Isotope Unlabeled Standard: d5-ethyl 3-methybutanoate ethyl 3-methybutanoate CAS: N/A 108-64-4 Mfg/Reference: synthesized Aldrich, Milwaukee, WI No.: ISO-74 324 % Purity (by GC-FID): N/A 87.4% Standard Curve


unlabeled 115

labeled 117 area ratio

11.361 3350867 16124 207.819

5.681 1822182 10650 171.097

2.840 1043567 7783 134.083

1.136 545127 4289 127.099

slope Rf

8.2107 0.122

y = 8.2107x + 116.88R² = 0.9757

0

50

100

150

200

250

0 2 4 6 8 10 12Are

a ra

tio

(io

n 1

15

/io

n 1

17

)


212

Response Factor of d2-isobutanol

Isotope Unlabeled Standard: d2-isobutanol isobutanol CAS: N/A 78-83-1 Mfg/Reference: synthesized Fisher, Fair Lawn, NJ No.: ISO-79 196 % Purity (by GC-FID): N/A 99.3% Standard Curve


unlabeled 74

labeled 76

area ratio

2.421 1387358 345867 4.011

0.968 687684 339327 2.027

0.468 432091 410322 1.053

0.234 197861 305494 0.648

0.094 159254 310886 0.512

slope Rf

1.5255 0.656

y = 1.5255x + 0.3734R² = 0.995

0

1

2

3

4

5

0.0 0.5 1.0 1.5 2.0 2.5 3.0Are

a ra

tio

(io

n 7

4/i

on

76

)


213

Response Factor of d2-isoamyl acetate

Isotope Unlabeled Standard: d2-isoamyl acetate isoamyl acetate CAS: N/A 123-92-2 Mfg/Reference: synthesized Aldrich, Milwaukee, WI No.: ISO-83 74 % Purity (by GC-FID): N/A 90.3% Standard Curve


unlabeled 70


0.105 104728 349266 0.300

0.262 772117 375522 2.056

0.525 1145880 642027 1.785

1.049 1866231 422745 4.415

2.623 1773329 113440 15.632

slope Rf

6.0101 0.166 y = 6.0101x - 0.649

R² = 0.976

0

5

10

15

20

0.0 0.5 1.0 1.5 2.0 2.5 3.0

Are

a ra

tio

(io

n 7

0/i

on

72

)


214

Response Factor of d5-ethyl pentanoate

Isotope Unlabeled Standard: d5-ethyl pentanoate ethyl pentanoate CAS: N/A 539-82-2 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: N/A 296 % Purity (by GC-FID): N/A 99.9 Standard Curve


unlabeled 88


4.728 1622578 282564 5.742

1.857 1412594 637979 2.214

0.929 1285894 1241451 1.036

0.464 648668 1336336 0.485

0.186 256265 1399108 0.183

slope Rf

1.2309 0.812

y = 1.2299x - 0.076R² = 0.9999

0

1

2

3

4

5

6

7

0 1 2 3 4 5

Are

a ra

tio

(io

n 8

8/i

on

93

)


215

Response Factor of d2-3-methyl-1-butanol

Isotope Unlabeled Standard: d2-3-methyl-1-butanol 3-methyl-1-butanol CAS: N/A 123-51-3 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: ISO-148 1058 % Purity (by GC-FID): N/A 78.1% Standard Curve


unlabeled 70

labeled 72

area ratio

1.161 11679784 6029549 1.937

0.464 5297421 6442510 0.822

0.231 3335765 5899612 0.565

0.116 3250796 5559691 0.585

0.046 1551266 5537305 0.280

slope Rf

1.4071 0.711

y = 1.4071x + 0.2699R² = 0.9768

0.0

0.5

1.0

1.5

2.0

2.5

0.0 0.5 1.0 1.5

Are

a ra

tio

(io

n 7

0/i

on

72

)


216

Response Factor of d2-2-methyl-1-butanol

Isotope Unlabeled Standard: d2-2-methyl-1-butanol 2-methyl-1-butanol CAS: N/A 137-32-6 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: N/A 590 % Purity (by GC-FID): N/A 95.8% Standard Curve


unlabeled 70

labeled 72

area ratio

2.975 8199139 1259561 6.510

1.190 3848920 1707206 2.255

0.595 1899249 1654228 1.148

0.297 937157 1241201 0.755

0.119 444565 1578854 0.282

slope Rf

2.1728 0.478

y = 2.1728x - 0.0596R² = 0.995

0

1

2

3

4

5

6

7

0 1 2 3 4Are

a ra

tio

(io

n 7

0/i

on

72

)


217

Response Factor of d11-ethyl hexanoate

Isotope Unlabeled Standard: d11-ethyl hexanotate ethyl hexanoate CAS: N/A 123-66-0 Mfg/Reference: CDN Aldrich, Milwaukee, WI No.: ISO-21 287 % Purity (by GC-FID): N/A 99.3% Standard Curve


unlabeled 99


0.197 146350 1622266 0.090

0.492 228891 1046768 0.219

0.983 645393 1282753 0.503

1.967 659733 778491 0.847

4.917 949068 389313 2.438

slope Rf

0.4957 2.018

y = 0.4957x - 0.0286R² = 0.9964

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0 1 2 3 4 5 6

Are

a ra

tio

(io

n 9

9/i

on

11

0)


218

Response Factor of d4-ethyl octanoate

Isotope Unlabeled Standard: d2-ethyl 2-methyl propanoate ethyl 2-methyl propanoate CAS: N/A 106-32-1 Mfg/Reference: synthesized Aldrich, Milwaukee, WI No.: ISO-22 288 % Purity (by GC-FID): N/A 98.4% Standard Curve


unlabeled 127

labeled 131

area ratio

5.170 4110447 471356 8.720

2.068 1947214 585986 3.323

1.034 966212 597518 1.617

0.517 447574 539078 0.830

0.207 214758 628811 0.342

slope Rf

1.6942 0.580

y = 1.6942x - 0.0814R² = 0.9995

0

2

4

6

8

10

0 2 4 6Are

a ra

tio

(io

n 1

27

/io

n 1

31

)


219

Response Factor of acetic acid

Standard: acetic acid CAS: 64-19-7 Mfg/Reference: SAFC, St. Louis, MO No.: 1099 % Purity (by GC-FID): 99.9% Standard Curve

acetaldehyde (density g/mL) 1.049

retention time (min) 16.2

Concentration (ug/mL) Area

2uL 1:100 dilution 40ABV 20.98 6.9

5uL 1:100 dilution 40ABV 52.45 13.6

10uL 1:100 dilution 40ABV 104.9 41

2uL 1:10 dilution 40ABV 209.8 106.9

5uL 1:10 dilution 40ABV 524.5 371.4

10uL 1:10 dilution 40ABV 1049 968.1

2uL 40ABV 2098 2806.9

5uL 40ABV 5245 7247

10uL 40ABV 10490 13843.5

slope Rf

1.3468 0.743

y = 1.3468x - 139.4R² = 0.9984

0

2000

4000

6000

8000

10000

12000

14000

16000

0 2000 4000 6000 8000 10000 12000

Are

a C

ou

nt

Concentration (ug/mL)

220

Response Factor of d4-β-damascenone

Isotope Unlabeled Standard: d4-β-damascenone β-damascenone CAS: N/A 23696-85-7 Mfg/Reference: synthesized Firmenich, Switzerland No.: ISO-1085 1085 % Purity (by GC-FID): N/A 89.3% Standard Curve


unlabeled 69


0.073 80790 313914 0.257

0.181 327220 464352 0.705

0.363 870785 672145 1.296

0.726 1270478 397271 3.198

1.815 412918 59906 6.893

slope Rf

3.8318 0.261

y = 3.8318x + 0.0497R² = 0.9941

0

1

2

3

4

5

6

7

8

0 1 1 2 2

Are

a ra

tio

(io

n 6

9/i

on

73

)


221

Response Factor of d3-guaiacol

Isotope Unlabeled Standard: d3-guaiacol guaiacol CAS: N/A 90-05-1 Mfg/Reference: CDN Sigma-Aldrich, St. Louis, MO No.: ISO-9 617 % Purity (by GC-FID): N/A 99.6% Standard Curve


unlabeled 124


0.296 43530 139601 0.312

0.741 68138 122602 0.556

1.481 448341 339782 1.319

2.962 819010 313879 2.609

7.406 713461 132557 5.382

slope Rf

0.7165 1.384

y = 0.7165x + 0.1893R² = 0.9918

0

1

2

3

4

5

6

0 2 4 6 8Are

a ra

tio

(io

n 1

24

/io

n 1

27

)


222

Response Factor of d2-trans-whiskey lactone

Isotope Unlabeled Standard: d2-trans-whiskey lactone trans-whiskey lactone CAS: N/A 39212-23-2 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: ISO-130 666 % Purity (by GC-FID): N/A 51.5% Standard Curve


unlabeled 99


0.064 77609 290225 0.267

0.161 206812 408854 0.506

0.322 438316 307494 1.425

0.645 1335543 490995 2.720

1.612 684362 142503 4.802

slope Rf

2.9202 0.342

y = 2.9202x + 0.3056R² = 0.9653

0

1

2

3

4

5

6

0.0 0.5 1.0 1.5 2.0Are

a ra

tio

(io

n 9

9/i

on

10

1)


223

Response Factor of 13C2-2-phenethyl alcohol

Isotope Unlabeled Standard: 13C2-2-phenethyl alcohol 2-phenethyl alcohol CAS: N/A 60-12-8 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: ISO-33 377 % Purity (by GC-FID): N/A 98.7% Standard Curve


unlabeled 91


5.848 17041004 3169702 5.376

2.339 7863291 3153643 2.493

1.170 3874655 2832114 1.368

0.585 1891567 2724069 0.694

0.234 864877 2784859 0.311

slope Rf

0.8929 1.177

y = 0.8929x + 0.2315R² = 0.9961

0

1

2

3

4

5

6

0 2 4 6 8Are

a ra

tio

(io

n 9

1/i

on

93

)


224

Response Factor of d2-cis-whiskey lactone

Isotope Unlabeled Standard: d2-trans-whiskey lactone trans-whiskey lactone CAS: N/A 39212-23-2 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: ISO-130 666 % Purity (by GC-FID): N/A 46.5% Standard Curve


unlabeled 99


0.221 53757 219739 0.245

0.552 118047 312803 0.377

1.104 218836 288048 0.760

2.209 924270 355450 2.600

5.522 554684 116630 4.756

slope Rf

0.7080 1.366

y = 0.708x + 0.0511R² = 0.9674

0

1

2

3

4

5

6

0 2 4 6 8Are

a ra

tio

(io

n 9

9/i

on

10

1)


225

Response Factor of d3-4-methylguaiacol

Isotope Unlabeled Standard: d3-4-methylguaiacol 4-methylguaiacol CAS: N/A 93-51-6 Mfg/Reference: synthesized SAFC, St. Louis, MO No.: ISO-117 644 % Purity (by GC-FID): N/A 100% Standard Curve


unlabeled 123


5.086 6896091 548017 12.584

2.034 8727925 1887058 4.625

1.017 6783094 2918659 2.324

0.509 4625970 3572590 1.295

0.203 2191429 3778477 0.580

slope Rf

2.4659 0.393

y = 2.4659x - 0.083R² = 0.9983

0

2

4

6

8

10

12

14

0 2 4 6Are

a ra

tio

(io

n 1

23

/io

n 1

25

)


226

Response Factor of d5-4-ethylguaiacol

Isotope Unlabeled Standard: d2-4-ethylguaiacol 4-ethylguaiacol CAS: N/A 2785-899 Mfg/Reference: synthesized SAFC, St. Louis, MO No.: ISO-71 618 % Purity (by GC-FID): N/A 98.1% Standard Curve


unlabeled 152


4.985 4992823 761451 6.557

1.994 6701591 2422162 2.767

0.997 4874989 3654887 1.334

0.498 3394426 4863630 0.698

0.199 1372047 4761892 0.288

slope Rf

1.3107 0.763

y = 1.3107x + 0.0551R² = 0.9995

0

1

2

3

4

5

6

7

0 2 4 6Are

a ra

tio

(io

n 1

52

/io

n 1

57

)


227

Response Factor of d3-p-cresol

Isotope Unlabeled Standard: d3-p-cresol d3-p-cresol CAS: N/A 106-44-5 Mfg/Reference: CDN Aldrich, Milwaukee, WI No.: ISO-5 425 % Purity (by GC-FID): N/A 98.5% Standard Curve


unlabeled 108

labeled 111

area ratio

5.196 1307609 176601 7.404

2.078 1719412 474785 3.621

1.039 1249138 861544 1.450

0.520 801677 1019327 0.786

0.208 496526 1162991 0.427

slope Rf

1..4209 0.704

y = 1.4209x + 0.1685R² = 0.9902

012345678

0 2 4 6Are

a ra

tio

(io

n 1

08

/io

n 1

11

)


228

Response Factor of d8-m-cresol

Isotope Unlabeled Standard: d8-m-cresol m-cresol CAS: N/A 108-39-4 Mfg/Reference: Aldrich SAFC, St. Louis, MO No.: ISO-4 539 % Purity (by GC-FID): N/A 93.8% Standard Curve


unlabeled 108

labeled 115

area ratio

4.742 21686876 2048880 10.585

1.897 29764034 9542716 3.119

0.948 20467879 11170228 1.832

0.474 14279977 15836671 0.902

0.190 6288968 14966468 0.420

slope Rf

2.2403 0.446

y = 2.2403x - 0.3257R² = 0.9876

0

2

4

6

8

10

12

0 1 2 3 4 5

Are

a ra

tio

(io

n 1

08

/io

n 1

15

)


229

Response Factor of d3-eugenol

Isotope Unlabeled Standard: d3-eugenol eugenol CAS: N/A 97-53-0 Mfg/Reference: synthesized Aldrich, Milwaukee, WI No.: ISO-36 640 % Purity (by GC-FID): N/A 98.1% Standard Curve


unlabeled 164


0.168 6794 55013 0.123

0.419 23642 41945 0.564

0.838 96818 94054 1.029

1.675 123948 75390 1.644

4.188 114206 22653 5.042

slope Rf

1.3155 0.760

y = 1.3155x - 0.0665R² = 0.9921

0

1

2

3

4

5

6

0 1 2 3 4 5Are

a ra

tio

(io

n 1

64

/io

n 1

67

)


230

Response Factor of d3-syringol

Isotope Unlabeled Standard: d2-ethyl 2-methyl propanoate ethyl 2-methyl propanoate CAS: N/A 91-10-1 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: ISO-60 611 % Purity (by GC-FID): N/A N/A Standard Curve


unlabeled 154

labeled 157

area ratio

3.284 3381619 817384 4.137

2.190 2090506 739854 2.826

1.095 608543 446841 1.362

0.547 1009339 1534358 0.658

0.365 711995 1272803 0.559

slope Rf

1.2682 0.789

y = 1.2682xR² = 0.9988

0112233445

0 1 2 3 4

Are

a ra

tio

(io

n 1

54

/io

n1

57

)


231

Response Factor of d3-(E)-isoeugenol

Isotope Unlabeled Standard: d2-(E)-isoeugenol (E)-isoeugenol CAS: N/A 97-54-1 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: ISO-43 183 % Purity (by GC-FID): N/A 92.5% Standard Curve


unlabeled 164

labeled 167

area ratio

4.731 1502212 88599 16.955

1.892 1190469 169018 7.043

0.946 1091538 311041 3.509

0.473 871157 462533 1.883

0.189 259738 274291 0.947

slope Rf

3.5374 0.283

y = 3.5374x + 0.2438R² = 0.9999

02468

1012141618

0 1 2 3 4 5Are

a ra

tio

(io

n 1

64

/io

n 1

67

)


232

Response Factor of ethyl vanillin

Isotope Unlabeled Standard: d3-vanillin ethyl vanillin CAS: N/A 121-32-4 Mfg/Reference: synthesized Sigma-Aldrich, St. Louis, MO No.: ISO-32 475 % Purity (by GC-FID): N/A 99.9% Standard Curve


unlabeled 166

labeled 155

area ratio

4.695 5561105 525516 10.582

1.878 6728710 1650289 4.077

0.939 3234969 2226836 1.453

0.470 2588756 3095085 0.836

0.188 994895 3218473 0.309

slope Rf

2.3172 0.416

y = 2.3172x - 0.3346R² = 0.9971

0

2

4

6

8

10

12

0 1 2 3 4 5Are

a ra

tio

(io

n 1

66

/io

n 1

55

)


233

Response Factor of d3-vanillin

Isotope Unlabeled Standard: d3-vanillin vanillin CAS: N/A 121-33-5 Mfg/Reference: synthesized SAFC, St. Louis, MO No.: ISO-32 70 % Purity (by GC-FID): N/A 84.1% Standard Curve


unlabeled 152


0.053 90007 581323 0.155

0.132 256913 498072 0.516

0.264 986398 914371 1.079

0.528 409299 167430 2.445

1.320 430856 75062 5.740

slope Rf

4.4167 0.229

y = 4.4167x - 0.0425R² = 0.9985

0

1

2

3

4

5

6

7

0.0 0.5 1.0 1.5

Are

a ra

tio

(io

n 1

52

/io

n 1

55

)


234

Response Factor of d5-ethyl vanillate

Isotope Unlabeled Standard: d3-ethyl vanillate ethyl vanillate CAS: N/A 617-05-0 Mfg/Reference: synthesized Alfa Aeasar, Lancaster, UK No.: ISO-73 1123 % Purity (by GC-FID): N/A 99.9%

Standard Curve selected ion mass ratio

unlabeled 196


0.038 30912 387496 0.080

0.095 45802 234305 0.195

0.190 155963 448551 0.348

0.381 263567 240807 1.095

0.952 183540 76150 2.410

slope Rf

2.6071 0.380

y = 2.6071x - 0.038R² = 0.9911

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0 0.2 0.4 0.6 0.8 1.0Are

a ra

tio

(io

n 1

96

/io

n 2

01

)


235

Response Factor of d3-syringaldehyde

Isotope Unlabeled Standard: d3-syringaldehyde syringaldehyde CAS: N/A 134-96-3 Mfg/Reference: synthesized SAFC, St. Louis, MO No.: ISO-78 1065 % Purity (by GC-FID): N/A 99.9% Standard Curve


unlabeled 182


21.789 8081107 236287 34.200

8.716 3621510 307692 11.770

4.358 1835874 323291 5.679

2.179 821049 315918 2.599

0.872 364613 359941 1.013

slope Rf

1.5988 0.603

y = 1.5988x - 1.0709R² = 0.9974

0

5

10

15

20

25

30

35

40

0 5 10 15 20 25

Are

a ra

tio

(io

n 1

82

/io

n 1

85

)


236

Appendix B: IRB approval letter for threshold determination in alcoholic matrices

237

Appendix C: Determination of odor thresholds in a 40% ABV ethanolic matrix

Determination of odor threshold of isoeugenol

Odor threshold: Isoeugenol added to a blank 40% ABV matrix

Procedure: ASTM Practice E679

Presentation: three Teflon plastic sniff bottles (two identical controls and one bottle containing the

added compound). Weakest concentrations were presented first

Number of scale steps: seven –each panelist observed each sample twice

Dilution factor per step: three

Temperature: samples were at room temperature (21°C)

Panelist selection: lab members experienced with sensory evaluation

Type of threshold: detection

Best-estimate threshold:

BET = 1860 μg/L

Log10BET= 3.27

Log standard deviation= 0.35

Concentration of isoeugenol in sample (μg/L) Individual Threshold

Log10 of individual thresholds Panelist 3.059 9.177 27.53 82.59 247.8 743.3 2230

A-1 0 1 1 0 0 0 0 3862 3.59

B-1 1 0 1 0 0 0 1 1287 3.11

C-1 0 0 0 0 0 0 1 1287 3.11

D-1 0 1 1 0 0 0 1 1287 3.11

E-1 1 0 1 0 0 0 0 3862 3.59

F-1 0 1 0 0 0 0 1 1287 3.11

G-1 1 0 1 0 0 0 1 1287 3.11

H-1 0 1 1 0 0 1 1 429 2.63

A-2 0 1 1 1 0 0 0 3862 3.59

B-2 0 0 0 0 0 1 1 429 2.63

C-2 0 0 1 0 1 0 0 3862 3.59

D-2 0 0 1 0 0 0 0 3862 3.59

E-2 0 0 1 0 1 0 0 3862 3.59

F-2 0 0 1 0 0 0 0 3862 3.59

G-2 0 0 1 0 0 0 1

1287 3.11

Ʃlog 10 49.03

Ʃlog 10/ replications 3.27

Group BET Threshold (μg/L) 1860

Standard Deviation (of Log10 Values) 0.35

238

Determination of odor threshold of ethyl vanillate

Odor threshold: ethyl vanillate added to a blank 40% ABV matrix










BET = 769000 μg/L

Log10BET= 5.90


Concentration of ethyl vanillate in sample (ug/L)

Individual Threshold

Log10 of individual thresholds Panelist 13660 40979 122938 368815 1106444 3319333 9958000

A-1 1 0 1 0 1 1 1 638806 5.81

B-1 0 1 0 1 1 1 1 212935 5.33

C-1 1 1 1 1 1 1 1 7886 3.90

D-1 1 1 0 0 1 0 1 5749254 6.76

E-1 1 1 1 1 1 1 1 7886 3.90

F-1 0 0 0 0 0 0 1 5749254 6.76

G-1 1 1 1 1 0 1 1 1916418 6.28

H-1 0 0 0 0 0 0 1 5749254 6.76

I-1 0 0 0 0 0 1 1 1916418 6.28

J-1 0 1 1 1 1 1 1 23659 4.37

A-2 0 1 1 1 1 0 1 5749254 6.76

B-2 1 1 1 1 0 1 0 17247762 7.24

C-2 1 1 1 0 1 1 1 638806 5.81

D-2 1 1 1 1 1 0 0 17247762 7.24

E-2 0 1 1 1 1 1 1 23659 4.37

F-2 0 0 1 0 0 0 0 17247762 7.24

G-2 1 0 1 1 1 1 1 70978 4.85

H-2 1 0 1 0 0 1 0 17247762 7.24

I-2 1 1 1 0 0 0 0 17247762 7.24

J-2 1 1 1 1 1 1 1 7886 3.90

Ʃlog 10 118.02

Ʃlog 10/replications 5.90

Group BET Threshold 796000

Standard Deviation 1.26

239

Determination of odor threshold of syringaldehyde

Odor threshold: syringaldehyde added to a blank 40% ABV matrix










BET = 6490000 μg/L

Log10BET= 6.81


Concentration of syringaldehyde in sample (ug/L) Individual Threshold

Log10 of individual thresholds Panelist 26700 80099 240296 720889 2162667 6488000 19464000

A-1 0 1 0 1 0 1 1 3745849 6.57

B-1 0 0 0 0 1 1 1 1248616 6.10

C-1 1 1 0 0 1 1 1 1248616 6.10

D-1 0 0 0 0 1 0 0 33712637 7.53

E-1 0 0 0 1 1 0 0 33712637 7.53

F-1 0 1 1 0 1 0 0 33712637 7.53

G-1 0 1 1 0 0 0 0 33712637 7.53

H-1 0 1 1 0 0 0 0 33712637 7.53

I-1 0 0 0 1 1 1 1 416205 5.62

A-2 0 1 0 0 0 0 1 11237546 7.05

B-2 0 0 0 1 1 1 1 416205 5.62

C-2 1 0 0 0 0 0 1 11237546 7.05

D-2 0 1 0 1 0 0 0 33712637 7.53

E-2 1 0 1 0 1 1 1 1248616 6.10

F-2 1 0 0 0 0 1 0 33712637 7.53

G-2 1 1 0 0 0 0 0 33712637 7.53

H-2 0 1 0 0 1 1 1 1248616 6.10

I-2 0 0 0 0 1 1 1 1248616 6.10

Ʃlog 10 122.62

Ʃlog 10/replications 6.81

Group BET Threshold 6490000

Standard Deviation 0.75

240

Appendix D: IRB approval letter for descriptive analysis panels

241

Appendix E: Rum DA panel recruitment questionnaire

Rum Descriptive Analysis Panel

Please answer all the questions below.

Q1

Please enter your first name

Q2

Please enter your last name

Q3

Please enter your e-mail address

Q4

Do you have any food or beverage allergies?

• Yes

• No

Q5

If yes, please list your allergy/sensitivity

Q6

Are you 21 years old or older?

• Yes

• No

Q7

Do you enjoy drinking distilled spirits?

• Yes

• No

Q8

If yes, how do you prefer them

• Neat

• With water

• On the rocks

242

• With a mixer

• Other

•

Q9

How often do you drink distilled spirits?

• Everyday

• 4-6 Times a Week

• 2-3 Times a Week

• Once a Week

• 2-3 Times a Month

• Once a Month

• Once every couple months

• Once a year

• Only on special occasions

• Never

Q10

What types of distilled spirits do you typically drink? (Select all that apply)

• Brandy

• Bourbon

• Cognac

• Gin

• Rum

• Scotch

• Tequilla

• Vodka

• Whiskey

• Other

•

• None

Q11

Do you smoke?

• Yes

• No

243

Q12

Are you pregnant or breastfeeding?

• Yes

• No

Q13

Are you a native English speaker?

• Yes

• No

Q14

Have you ever suffered a chronic disease related to alcohol consumption?

• Yes

• No

Q15

Please check ALL times that you are available every day between ______ and ______.

• 10am - 11am

• 11am - 12pm

• 12pm - 1pm

• 1pm - 2pm

• 2pm - 3pm

• 3pm - 4pm

• 4pm - 5pm

• 4:30pm - 5:30pm

• 5pm - 6pm

244

Appendix F: Rum DA panel prescreening ballot

PRESCREENING TEST

Basic Tastes

Instructions:

1. Taste each solution in the order presented below. 2. For each solution identify which of the basic tastes (sweet, sour, salty, bitter) is being

presented. You can also answer non if no taste is perceived.

568 ___________________________________ 425 ___________________________________ 328 ___________________________________ 741 ___________________________________ 195 ___________________________________

Aroma Identification

Instructions:

1. Smell the samples in the order presented below. 2. To smell the sample, remove the lid from the cup and then take short shallow sniffs. 3. Next check if you perceive an odor, and then identify what the odor is (ie/grape). 4. For each solution identify which of the basic tastes (sweet, sour, salty, bitter) is being

presented. You can also answer non if no taste is perceived.

Odor Perceived Describe Odor

625 Yes No _____________________________________________________________________

539 Yes No

_____________________________________________________________________ 842 Yes No

_____________________________________________________________________ 654 Yes No

_____________________________________________________________________ 197 Yes No

_____________________________________________________________________

245

Appendix G: Rum DA panel informed consent form

INFORMED CONSENT FORM FOR SENSORY EVALUATION PANELISTS

“Descriptive Analysis Panel of Rum and Rum Model Systems”

You are invited to participate in a study involving sensory evaluation of rum. The goal of this research is to determine

the sensory characteristics of premium aged rums as well as the effect of ethanol on the sensory perception of rums.

Several types of rum and rum models will be evaluated in this study to determine the sensory similarities and

differences of various premium rums. The products will be evaluated using a sensory Descriptive Analysis panel. You

will be presented with rum samples and asked to generate terms, definitions and references to describe sensory

attributes. Panelists will work together to define the significant attributes and corresponding references. Once

determined, the significant attributes will be rated individually based on the references. This research will allow the

investigators to gain a better understanding of the sensory attributes associated with premium aged rums.

To participate in this study, you must first complete a sensory screening process involving the identification of odors

and basic taste solutions. Based on the results of this sensory screening you may not qualify to participate in this study.

Screening and testing for this study will be conducted in Agricultural Bioprocess Laboratory (ABL) Room 201 and

Bevier Hall Room 376. We anticipate that the research will consist of 5 hours of testing sessions per week for 8-weeks

for a total of 38 hours of testing. Each day testing will last approximately 1 hour, either a one hour session or two 30

minute session. The total number of sessions required for each panelist is 52 (24 one hour sessions and 28 thirty minute

sessions), and you will be compensated monetarily for your participation in the amount of $99 upon completion of the

study. Participation in this study is voluntary and you are free to withdraw from the study at any time for any reason

and it will have no effect on your grades at, status at, or future relations with the University of Illinois. The

experimenter(s) also reserve the right to terminate the participation of an individual subject at any time. You will be

terminated if you miss sessions, are consistently late, cannot follow directions or become intoxicated after evaluating

samples. In the event you withdraw from, or are terminated from the study by the experimenter you will be

compensated at a rate of $2.50 per hour of testing completed.

A complete list of the rums will be available for review after testing has concluded. Information about the references

will be available throughout the panel. If you have any food or beverage allergies you should not participate in this

research. All foods and ingredients have been designated as safe for food use by their respective manufacturers and

are commonly found in commercially available foods.

During this study there are slight physical risks associated with alcohol consumption. These risks will be minimized

since all rum samples tasted will be expectorated. In addition you will be asked to eat something one hour before

coming to panel. Even if some rum is swallowed during tasting, the total amount of alcohol you receive at each

session will be similar to that of a 12 oz beer and should not be enough to intoxicate you. Additionally, you should

not drive yourself to or from the testing sessions. You must be over 21 years of age to participate. Anyone who has a

chronic disease related to alcohol consumption or is pregnant or breastfeeding should not participate in this study.

The University of Illinois does not provide medical or hospitalization insurance coverage for participants in this

research study nor will the University of Illinois provide compensation for any injury sustained as a result of

participation in this research study, except as required by law.

Your participation in this study will be kept confidential, but not always. In general, we will not tell anyone any

information about you. When this research is discussed or published, no one will know that you were in the

study. However, laws and university rules might require us to tell certain people about you. For example, your

records from this research may be seen or copied by the following people or groups: Representatives of the

246

university committee and office that reviews and approves research studies, the Institutional Review Board (IRB)

and Office for Protection of Research Subjects; other representatives of the state and university responsible for

ethical, regulatory, or financial oversight of research; federal government regulatory agencies such as the Office of

Human Research Protections in the Department of Health and Human Services. Any publications or presentations of

the results of the research will only include information about group performance. Data gathered from the entire

project will be summarized in the aggregate, excluding references to any individual responses. Photos of the

panelists participating in this research may be taken and used in oral presentations, in order to give information

about the experiment procedure. Names of panelists will not be associated with the photos. Panelists may opt for

not having their photographs taken and this option is included on the consent form. The aggregated results of our

analysis will be for journal articles and conference presentations. Again, your input is very important to us and any

information we receive from you will be kept secure and confidential.

You are encouraged to ask any questions about this study whether before, during, or after your participation. However,

specific questions about the samples that could influence the outcome of the study will be deferred to the end of the

experiment. Questions can be addressed to Dr. Keith Cadwallader (217-333-5803, [email protected]) or Chelsea

Ickes (443-845-1404, [email protected]). If you have any questions about your rights as a participant in this study

or any concerns or complaints, please contact the University of Illinois Office for the Protection of Research Subjects

(OPRS) at 217-333-2670 or via email at [email protected].

I understand the above information and voluntarily consent to participate in the study described above.

I have been offered a copy of this consent form. Yes No

I am 21 years of age or older. Yes No

I agree to have photographs taken of me while participating in this research. Yes No

I am not pregnant or breastfeeding . Yes No

I do not take any medication. Yes No

I do not have any chronic diseases related to alcohol consumption. Yes No

I do not smoke. Yes No

Signature Date

Print Name

https://webmail.illinois.edu/owa/[email protected]/redir.aspx?SURL=TqTDwj5nHNNppe0W5FlNslIut853g5D6XE2SxU3eAAsWmv86C6_SCGgAdAB0AHAAOgAvAC8AaQBsAGwAaQBuAG8AaQBzAC4AZQBkAHUALwBkAHMALwBkAGUAdABhAGkAbAA_AGQAZQBwAGEAcgB0AG0AZQBuAHQASQBkAD0AaQBsAGwAaQBuAG8AaQBzAC4AZQBkAHUATgBFADMANAA0ACYAcwBlAGEAcgBjAGgAXwB0AHkAcABlAD0AYQBsAGwAJgBzAGsAaQBuAEkAZAA9ADAAJgBzAHUAYgA9AA..&URL=http%3a%2f%2fillinois.edu%2fds%2fdetail%3fdepartmentId%3dillinois.eduNE344%26search_type%3dall%26skinId%3d0%26sub%3d

mailto:[email protected]

mailto:[email protected]

https://webmail.illinois.edu/owa/[email protected]/redir.aspx?SURL=VFuIcwKQbM7gNbaoqA32FSD4InFxhkLJfU8h7STtqs8Wmv86C6_SCG0AYQBpAGwAdABvADoAaQByAGIAQABpAGwAbABpAG4AbwBpAHMALgBlAGQAdQA.&URL=mailto%3airb%40illinois.edu

247

Appendix H: Sample term generation sheet for descriptive analysis panels

Term Generation Day 1

Sample :XXX Terms Definitions References

Aroma

Aroma-by-

mouth

Texture/

Mouthfeel

Taste

Aftertaste

248

Appendix I: Product information for attributes for rum descriptive analysis panel

Modality Attribute Product Name Brand Manufacturer Location A

rom

a

alcohol Ethanol 190 Proof, USP Grade Decon Laboratories, Inc. Decon Labs, Inc. King of Prussia, PA

almond roasted almonds, unsalted True Goodness by Meijer Meijer Distribution Inc Grand Rapids, MI

brown spice

Ground nutmeg McCormick McCormick & Co., Inc. Hunt Valley, MD

brown sugar

Dark Brown Sugar Meijer Meijer Distribution Inc Grand Rapids, MI

butter Sweet Cream Butter Unsalted Meijer Meijer Distribution Inc Grand Rapids, MI

caramel Smucker's Sundae Syrup Caramel Smucker's The J.M. Smucker

Company Orrville, OH

cherry Cherry Pie Filling Meijer Meijer Distribution Inc Grand Rapids, MI

chocolate Unsweetened Baking Chocolate Bar Baker's Kraft Foods Group, Inc. Northfield, IL

citrus lime Susie

coconut Toasted Flake Coconut Coral Bay Marx Brothers, Inc. Birmingham AL

dried fruit Sunsweet Amazin Pruntes Pitted Sunsweet Sunsweet Growers Inc. Yuba City, CA

maple Maple Extract McCormick McCormick & Co., Inc. Hunt Valley, MD

phenolic Plastic Bandages Meijer Meijer Distribution Inc Grand Rapids, MI

roasted Brown Malt Thomas Fawcett & Sons

Ltd. The Country Malt Group

Castleford, West Yorkshire

smokey Hardwood Smoked Bacon John Morrell John Morrell & Co Cincinnati, OH

vanilla All Natural Pure Vanilla Extract McCormick McCormick & Co., Inc. Hunt Valley, MD

walnut raw chopped walnuts True Goodness by Meijer Meijer Distribution Inc Grand Rapids, MI

woody Oak wood chips LD Carlson Company LD Carlson Company Kent, OH

Mouthfeel

astringent Lipton Pure Green Tea Lipton Unilever Englewood Cliffs, NJ

slick Vegetable Glycerin Heritage Store Heritage Store Virginia Beach, VA

warming Ethanol 190 Proof, USP Grade Decon Laboratories, Inc. Decon Labs, Inc. King of Prussia, PA

249

Appendix I (cont.): Product information for attributes for rum descriptive analysis panel

Modality Attribute Product Name Brand Manufacturer Location

Taste bitter Caffeine Fisher Scientific Fisher Scientific Fair Lawn, NJ

Aft

ert

aste

bitter Caffeine Fisher Scientific Fisher Scientific Fair Lawn, NJ

brown spice





coffee Eight O'clock French Roast Coffee Eight O'clock Eight O'clock Coffee

Company Montcale, NJ

Aro

ma

By

Mo

uth


brown spice





coconut Toasted Flake Coconut Coral Bay Marx Brothers, Inc. Birmingham AL





smokey Liquid Smoke - Hickory Wright's B&G Foods, Inc. Parsippany, NJ


walnut raw chopped walnuts True Goodness by Meijer Meijer Distribution Inc Grand Rapids, MI

woody Oak wood chips LD Carlson Company LD Carlson Company Kent, OH

250

Appendix J: Eigenvalues for factor loading using covariance matrix

Factor Eigenvalue Proportion Cumulative

1 71.5924322 0.8836 0.8836

2 2.9243101 0.0361 0.9197

3 2.0241442 0.025 0.9447

4 1.6381673 0.0202 0.9649

5 1.5601022 0.0193 0.9842

6 0.7638361 0.0094 0.9936

7 0.3044791 0.0038 0.9973

8 0.2158511 0.0027 1

251

Appendix K: Principle component analysis factor correlations of significant attributes on principal component 1 (PC1) and 2 (PC2) for all nine rum samples in covariance matrix

Modality Attribute PC1 PC2

Aroma

Brown Sugar 0.97 0.01

Caramel 0.97 -0.22

Maple 0.93 -0.15

Vanilla 0.99 -0.06

Alcohol -0.95 0.08

Citrus -0.81 -0.36

Coconut 0.82 0.25

Roasted 0.41 -0.01

Smoky -0.61 0.31

Phenolic -0.94 0.18

Chocolate 0.97 -0.15

Mouthfeel Warming -0.84 0.02

Slick 0.59 0.75

Taste Bitter -0.88 0.05

Aftertaste Brown Spice 0.85 0.17

Caramel 0.99 0.07

Aroma-by-mouth

Caramel 0.97 0.15

Maple 0.99 0.03

Vanilla 0.96 0.26

Coconut 0.85 0.26

252

Appendix L: Product information for attributes for ethanol dilution descriptive analysis panel

Modality Attribute Product Name Brand Manufacturer Location A

rom

a






dark fruit Sunsweet Amazin Prunes Pitted Sunsweet Sunsweet Growers Inc. Yuba City, CA




toasted Jet-Puffed Marshmallows Jet Puffed - Kraft Kraft Foods Group, Northfield, IL

woody Oak wood chips LD Carlson Company Kent, OH

Mouthfeel

astringent Lipton Pure Green Tea Lipton Unilever Englewood Cliffs, NJ

silky Almond milk - Reduced Sugar,

Vanilla Silk Whitewave Foods Broomfield, CO

slick Vegetable Glycerin Heritage Store Heritage Store Virginia Beach, VA

warming Ethanol 190 Proof, USP Grade Decon Laboratories, Inc. Decon Labs, Inc. King of Prussia, PA

Taste bitter Caffeine Fisher Scientific Fisher Scientific Fair Lawn, NJ

sweet C&H Pure Cane Sugar C&H Domino Foods, Inc Yonkers, NY

Aro

ma

By

Mo

uth







253

Appendix L (cont.): Product information for attributes for ethanol dilution descriptive analysis panel

Modality Attribute Product Name Brand Manufacturer Location

Aft

ert

aste

bitter Caffeine Fisher Scientific Fisher Scientific Fair Lawn, NJ

brown spice




plastic Magnetic Vinyl Shower Curtain

Liner Maytex Maytex Mills, Inc China

254

Appendix M: Principle component analysis factor correlations of significant attributes on principal component 1 (PC1) and 2 (PC2) for DR12 dilution samples in covariance matrix


Aroma

Alcohol 0.98158 0.19104

Caramel 0.81937 0.57327

Maple 0.58324 0.8123

Vanilla 0.61463 0.78882

Roasted 0.99461 -0.10367

Woody 0.99281 0.11971

Mouthfeel

Astringent 0.99847 -0.05531

Silky -0.99597 0.08973

Slick 0.98852 0.15108

Warming 0.98994 -0.14151

Taste Bitter 0.97549 -0.22004

Aftertaste Bitter 0.97201 -0.23495

Brown Spice 0.95102 0.30913

Aroma-by-mouth

Alcohol 0.99663 -0.08202

Caramel 0.98095 0.19427

Maple 0.92943 0.36899

Vanilla 0.87614 0.48205

Woody 0.97612 -0.21724

255

Appendix N: Principle component analysis factor correlations of significant attributes on principal component 1 (PC1) and 2 (PC2) for RA7 dilution samples in covariance matrix


Aroma

Alcohol 0.99353 0.1136

Caramel 0.99382 0.11099

Vanilla 0.98308 0.18316

Dark Fruit 0.55935 0.82893

Mouthfeel

Astringent 0.99995 0.01046

Slick 1 -0.00066

Warming 0.98745 -0.15792

Taste Bitter 0.99315 -0.11687

Aftertaste

Bitter 0.99955 0.03008

Brown Spice 0.99952 -0.03092

Vanilla 0.97228 0.23381

Plastic 0.99901 0.04458

Aroma-by-mouth

Alcohol 0.99893 -0.04617

Caramel 0.98501 0.17248

Maple 0.99688 0.07897

Vanilla 0.96137 0.27525

256

Appendix O: Detailed day-by-day procedural description of rum descriptive analysis panel

Day 1: Introductory session and term generation for rum samples

After granting informed consent (Appendix G), the panel facilitator introduced themselves and their role in

the panel. Panelists then introduced themselves to each other. The panel facilitator gave a short presentation

that introduced basic sensory science principles and DA methodology. The panelists were then presented

with four of the twelve rums (Bacardi Gold, El Dorado 12 year, Ron Abuelo 7 year, and Model 2). Panelists

were instructed to evaluate the first sample and develop terms for the different modalities (aroma, aroma-by-

mouth, taste, texture/mouthfeel, and aftertaste), they were presented with a term generation worksheet

(Appendix H) to aid in the development of terms. The panelists discussed their findings briefly and then were

instructed to evaluate the remaining three samples. The panelists evaluated all four rums and were asked to

generate attributes for all modalities. Panelists were also encouraged to identify corresponding definitions and

references matching these attributes. During discussion, panelists mentioned all terms, definitions and

references that had been generated. Through discussion, the panelists compiled an extensive list of possible

attributes and corresponding references to be investigated the following day.

While rating, panelists were asked to determine and effective rinsing protocol. Rinses such as water (warm

(40°C) and room (23°C)), crackers (saltine (Great Value Unsalted Tops Saltine Crackers, Wal-Mart Stores,

Inc., Bentonville, AR) and water (Carter’s Water Crackers)) and bread (SaraLee Honey Wheat Bread, Bimbo

Bakeries USA, Inc., Horsham, PA) were made available. Panelists were given the opportunity to suggest any

other rinses they would like to evaluate.

The panel was then dismissed and panelists were instructed to return at the same time the following day.

Once the session concluded, the panel facilitator compiled the final list of terms, attributes and references,

including any that may have been written on a panelist’s term generation sheet, but not brought up during

panel discussion due to time restraints. The initial list consisted of 38 terms and 35 references. The panel

facilitator then purchased the references the panel had identified from local stores.

Day 2: Term generation and reference refinement for rums

After signing in, panelists were presented will all of the references they had requested during term generation

on day 1. Panelists determined if these references correctly represented the attribute detected in the rums in

terms of both quality and concentration. After examining all of the references, panelists were presented with

4 new rum samples (Bacardi White, Appleton Estate 12 Year, Dictador XO, and Model 3) to continue term

generation and reference refinement. After panelists had a chance to evaluate all of the rum samples, panelists

discussed their findings and put forth new terms and references. Panelists reached consensus on terms and

references that needed to be added, modified or eliminated for the next session. Panelists also selected their

257

rinse protocol, choosing bread, warm and then room temperature water. Panelists were then dismissed and

the facilitator compiled the observations from the session in preparation for the next day.


After signing in, panelists were presented with the references that they kept from day 2 and as well as the new

references they generated the previous day. Panelists were asked to evaluate the references and then evaluate

the four rum samples (Appleton Estate V/X, Diplomatico Reserva 12 year, Ron Zacapa and Model 1). At

the end, panelists were brought back together to discuss if the new references fit in terms of quality and

intensity and were given an opportunity to propose any new terms as well. Panelists reached consensus on

terms and references that needed to be added, modified or eliminated for the next session. Panelists were

then dismissed and the facilitator compiled the observations from the session in preparation for the next day.


After signing in, panelists were presented with the references that they kept from day 3 and the new

references that they generated the previous day. At the beginning of the session, panelists were introduced to

the rum aroma wheel. The wheel was presented to help aid in the generation of terms to describe the rum

samples. Panelists were asked to evaluate the references and then evaluate the four rum samples (Appleton

Estate V/X, Appleton Estate 12 year, Ron Abuelo 7 year, and Model 1) using the wheel to help aid in term

generation. At the end, panelists were brought back together to discuss if the new references fit in terms of

quality and intensity and were given an opportunity to propose any new terms as well. Panelists reached

consensus on terms and references that needed to be added, modified or eliminated for the next session.

Panelists were then dismissed and the facilitator compiled the observations from the session in preparation

for the next day.


After signing in, panelists were presented with the references that they kept from day 4 and the new

references that they generated the previous day. Panelists were again provided with the flavor wheel and

encouraged to use the wheel to aid in the evaluation of the samples. Panelists were asked to evaluate the

references and then evaluate the four rum samples (Bacardi Gold, El Dorado 12 year, Dictador XO Insolent,

and Model 2) using the wheel to help aid in term generation. After evaluating the samples, group discussion

focused of going through the samples together and making sure everyone was able to pull out the same

attributes in each sample. Panelists aided each other in helping to pull out those terms in the samples by

checking the references and going back and evaluating the samples. Panelists also discussed terms that need

to be added, modified or eliminated for the next session. Panelists were then dismissed and the facilitator

compiled the observations from the session in preparation for the next day.

258

Day 6: Reference refinement for rums

After signing in panelists were presented with the references they had generated in previous sessions as well

as the new references they had asked for on day 5. After evaluating the references the panelists were asked to

evaluate the samples (Bacardi White, Diplomatico Reserva 12 Year, Ron Zacapa, Model 3). Then panelists

were asked to refine the list of terms, definitions and references to a semi-final list for which to start scaling

with. Panelists were asked to clarify terms and definitions that had not been made clear during previous panel

sessions. Panelists were also asked to clarify which modality they detected many of the aroma and aroma-by-

mouth attributes as they had been used fluidly between the two during term generation. Panelists were

informed that scaling would start the next day. The panel facilitator asked if it would be easier for panelists if

the terms were grouped into larger categories to aid in identifying and ranking those terms during scaling and

testing. The panelists agreed that would be helpful and the list was adjusted for the next day. The facilitator

then dismissed the panel and compiled the observations from the session to prepare for the next day. The

final list of generated terms (table X) consisted of 28 aroma, 25 aroma-by-mouth, 4 mouthfeel, 1 taste, and 5

aftertaste attributes, for a total of 63 initial attributes that were identified in the rum samples.

Day 7: Introduction to scaling method and scaling of aroma attributes

After signing in, the panelists were also informed that a subset of the samples, the models, were being

removed from the study due to the fact that they were clearly distinguishable from the rums they were models

of. Continued evaluation of the samples at this point in time would be futile until further modifications to the

model system could be made.

The panelists were then introduced to the concept of scaling. Panelists were instructed on the use of a 15

point categorical scale with 0 being no perception of an attribute and 15 being the strongest intensity of that

attribute in the rum samples. Panelists were then trained how to rate the references. Panelists were instructed

to select the sample with the highest intensity of that attribute and assigning it a score of 15 and then rate the

reference and other two according to where they fell along that 15 point scale. Panelist then used the three

rum samples (Appleton Estate V/X, Diplomatico Reserva 12 Year, and Dictador XO Insolent) to score the

references for the 35 aroma terms the panel had generated. Due to the number of samples and references that

the panelists had to rate, there was no time for group discussion and panelists were dismissed once they

finished rating the references. After the panel, the reference and sample scores were calculated and any

reference that received a score of 15 or greater was adjusted to be re-rated the following day.

Day 8: Scaling of aroma attributes

After signing in, the same or modified references (where necessary) were presented to the panelists for

continued rating and scoring. Panelists were also presented with reference scores for the unmodified

259

references and their individual and group score from the previous session to aid in training. Panelists were

then presented with three rum samples (Bacardi Gold, Appleton Estate 12 year, and Ron Abuelo 7 year) and

asked to rate the aroma references that had been adjusted from the previous day. Panelists were encouraged

to try different evaluation protocols, such as letting the sample air for 3 seconds before evaluating, if the

reference intensity was to strong when the sample was opened and evaluated immediately. Panelists were

asked to make note if a different evaluation procedure brought the reference intensity on scale. Panelists

were also asked to practice scoring the rums for the attributes with established reference ratings, using the

reference rating as the anchor to score the rum samples, if they had time. After the panel, the reference and

sample scores were calculated and any reference that received a score of 15 or greater was adjusted to be re-

rated the following day.

Day 9: Scaling of aroma atrributes

After signing in, the same or modified references (where necessary) were presented to the panelists for

continued rating and scoring. Panelists were also presented with reference scores for the unmodified

references and their individual and group score from the previous session to aid in training. The panelists

were presented with three samples (Bacardi White, El Dorado 12 year, and Ron Zacapa) and asked to score

the references that had been adjusted from the previous day. Due to the intensity of some of the references

in comparison to the rum samples, two evaluation protocols were developed. The first protocol was to

remove the lid of the reference and immediately evaluate the aroma intensity using bunny sniffs. The second

protocol was to remove, the lid, let the sample air for three seconds and then evaluate the aroma intensity.

Panelists were free to provide feedback as to how references, definition or evaluation procedures may need to

be adjusted. The facilitator then dismissed the panel and compiled the reference and samples score to provide

feedback to the panelists the following day.

Day 10: Scaling Mouthfeel and Taste attributes

After signing in, the panelists were presented with all of the references for the mouthfeel and taste modalities

for reference scoring. Panelists were asked to score the samples in the same way they had scored the aroma

attributes. Panelists were reminded that all references needed to be evaluated in the mouth. Panelists were

then presented with three rum samples (Bacardi Gold, Appleton Estate 12 year, and Dictador XO Insolent)

and asked to rate the 10 mouthfeel, taste, and aftertaste references the panel had generated during term

generation. Panelists were also asked to practice scoring the rums for the attributes with established reference

ratings, using the reference rating as the anchor to score the rum samples. Panelists were free to suggest any

terms, definitions or references that needed to be modified or eliminated. Panelists chose to remove tingling

from the mouthfeel terms. The facilitator then dismissed the panel. After the panel, the reference and sample

scores were calculated and any reference that received a score of 15 or greater was adjusted to be re-rated.

260

Day 11: Scaling aroma-by-mouth attributes

After signing in, panelists were presented with all of the aroma-by-mouth attributes for reference scoring.

Panelists were asked to score the samples in the same way they had scored the previous attributes. Panelists

were reminded of the importance of evaluating all of the reference in the mouth, as well as the importance of

rinse protocol. Panelists were then presented with three rum samples (Bacardi White, Diplomatico Reverva

12 year, and Ron Abuelo 7 year) and asked to rate the 23 aroma-by-mouth attributes the panel had generated

during term generation. Panelists were free to suggest any terms, definitions or references that needed to be

modified or eliminated. Panelists chose to remove citrus, green apple, butterscotch, and cucumber from the

aroma-by-mouth attributes. The facilitator then dismissed the panel. After the panel, the reference and

sample scores were calculated and any reference that received a score of 15 or greater was adjusted to be re-

rated the following day.

Day 12: Scaling

After signing in, panelists were presented with the same or modified references (where necessary) continued

rating and scoring. Panelists were provided with their individual scores along with the scores of the other

panelists and group average in order to aid in training. Panelists were then provided with a paper ballot similar

to the one they use during booth testing, both in terms of modality presentation order, grouping of terms,

and attribute and reference information. Panelists were instructed to rate the three samples (Appleton Estate

V/X, El Dorado 12 year, and Ron Zacapa), one at a time, using the reference scores that had been generated

during the scaling sessions the previous days. Panelists were asked to re-score any references that had been

modified from previous days and indicate their score on the references sheet. The facilitator dismissed the

panelist’s once they had finished rating their rums and informed them that they would receive the same

samples the following day to be able to discuss their ratings and work towards a better group consensus. The

facilitator then compiled the sample scores to provide feedback and aid in training the following day.

Day 13: Scaling

After signing in, panelists were presented with their references and the same samples from the previous day

(Appleton Estate V/X, El Dorado 12 year, and Ron Zacapa). Panelists were also provided with their scores

as well as the scores of the other panelists and the group average to aid in guiding them to rate the samples in

a more uniform way. Panelists were encouraged to rerate the samples using the references as anchors.

Panelists then discussed their rating and provided feedback as to any terms, definitions or references that

needed to be modified or eliminated. During the group discussion, panelists elected to form protocols for the

evaluation of each modality as follows: aroma – removed the cover and let air for three seconds without

fanning the sample, then place your nose in the middle of the glass and evaluate the aroma, aroma-by-mouth,

place the sample in your mouth and hold for 5 seconds and then evaluate the aroma-by-mouth attributes,

261

mouthfeel and taste – place the samples in your mouth and then hold for 3 seconds and evaluate the

attributes, aftertaste, place the samples in your mouth, hold for three seconds and expectorate and they

evaluate once the burning sensation has dissipated. Panelists also chose to eliminate raisin, green and vegetal

aroma terms. The facilitator then dismissed the panel and compiled the sample scores to provide feedback

and aid in training the following day.

Day 14: Scaling

After signing in, panelists were provided with their references and scores from the previous day. Panelists

then evaluated one sample (Diplomatico Reserva 12 year), as the previous day they had indicated being able

to focus on and discuss one sample would be most beneficial to aiding in reaching group consensus in rating

the rums. During discussion, panelists focused on attributes they had the most difficulty with including

baking spices, alcohol, and woody terms. Panelists elected to consolidate dried apricots and prunes into one

term, dried fruit. Panelists also eliminated the terms toffee, and liquorice, as well as removed coffee from

aroma and aroma-by-mouth. The facilitator then dismissed the panel and compiled the sample scores to

provide feedback and aid in training the following day.

Day 15: Individual practice booth testing

Panelists attended two 30 minutes booth testing sessions in order to become familiar with the individual

testing and Compusense software (session 1: Bacardi White and Appleton Estate 12 year, session 2: Dictador

XO Insolent and Bacardi Gold). When panelists arrived at the session, they were presented with a tray of

references, their rinses, and a table which detailed each term and its corresponding definition, reference and

intensity rating. Panelists were also provided with their scores from the previous day. Panelists were

encouraged to review their tray of references to refresh themselves as to the intensity and rating of the

references for each attribute. Panelists were allowed to review their references for as long as they wanted.

After reviewing all of the references, panelists moved to the booths to individually rate the samples.

Testing took place in individual booths, with positive airflow and temperature set to 68°F. All samples were

presented in black double old-fashion glasses (Threshold, Target, USA) labeled with random three digit codes

and evaluated under red lighting. The samples in each set were presented to the panelists in random order.

The ballot was presented to panelists using Compusense seven as the interface in the booth and Compusense

five to launch the test (Need Product Information). When panelists entered the booth they were instructed

to enter their panelist code, and then indicated to the session facilitator that they were ready for the test by

turning on their light. Panelists reviewed both samples at the same time and were told to check the sample

codes before proceeding with the test. Panelists were cued with instructions for each modality, and the test

had built in timers so that the sampling procedures were consistent between all panelists. Panelists strictly

followed the tasting protocols they developed (described in day 13). Panelists were given a 0 - 15 point

262

categorical scale with which to rate the samples, with 0 being no perception of the attribute and 15 being the

greatest intensity of the attribute in the rum samples. Panelists were provided with the attribute, definition,

reference and reference score for each term. After evaluating the first samples, panelists were given a 2

minute break where they were instructed to follow the rinse procedure to cleanse their pallet before the next

sample. After rating the second sample panelists indicated to the facilitator that they were done.

Panelists then came back a second time during the day, at least 30 minutes after finishing the first session to

evaluate a second sample set. Panelists were asked to review any references that they felt gave them trouble

but were told they did not have to review every attribute. The second sample was also evaluated in the

booths. After all panelists had finished the data was exported and evaluated in Microsoft excel.

Day 16: Scaling

After signing in, panelists were presented with their references and samples (Appleton Estate 12 year and

Dictador XO Insolent). Panelists were also provided with their scores as well as the scores of the other

panelists and the group average from the previous day to aid in guiding them to rate the samples in a more

uniform way. Panelists were asked to rate the samples using the references as anchors on the scale. The

facilitator then compiled all of the ratings in order to provide feedback and aid in group discussion. Panelists

decided to eliminate the cucumber aroma and brown sugar aroma-by-mouth terms. The facilitator then

dismissed the panel and compiled the sample scores to provide feedback and aid in training the following day.

Day 17: Scaling

After signing in, panelists were presented with their references, scores from the previous day and samples

(Bacardi White, Diplomatico Reserva 12 year, Appleton Estate V/X, and Ron Zacapa). The four samples

selected were chosen to give the broadest spectrum possible in sample variations for panelists to practice

with. Panelists were only presented with the 15 references that they had the most difficulty rating consistently.

Panelists were again reminded that at least one of the samples should receive a 15 for each attribute when

sampling. All of the scores were then tabulated and terms that had standard deviations of 3 or greater were

discussed. The facilitator encouraged panelists to finalize all terms, definition, reference scores and evaluation

protocols. During the discussion, panelists elected to eliminate clove, malty, and cola from aroma and aroma-

by-mouth terms, cooling from aroma, and to change the nutmeg term to brown spice. The facilitator then

dismissed the panel and compiled final list of all sample modalities, terms, definition and intensities in Table

6.1 for the rum samples. The final list consisted of 19 aroma, 12 aroma-by-mouth, 3 mouthfeel, 1 taste, and 5

aftertaste terms.

Day 18: Scaling

263

On the final day of scaling, panelists were provided with their finalized list of terms and reference and their

scores from the previous day to aid in training to score more consistently as a group. Panelists were instructed

on booth protocol and reminded of the specific sample evaluation procedures that they had set for each

modality. Panelists were then given one last opportunity to change any of the sampling protocols or the order

in which the modalities or attributes were presented on the ballot. Panelists then reviewed their references

and practiced rating one sample (El Dorado 12 year). Panelist discussed terms they were struggling with rating

consistently. The panelists were then dismissed and the panel facilitator compiled the observations from the

session in preparation for booth testing the next day.

Day 19: Individual booth testing of rum sample sets 1 and 2

Panelists attended two 30 minutes booth testing sessions, at least 30 minutes apart, to rate the attributes they

had identified in the rums according to the selected references (session 1: Bacardi Gold and Dictador XO

Insolent, session2: Appleton Estate V/X and Ron Zacapa). When panelists arrived at the session, they were

presented with a tray of references, their rinses, and a table which detailed each term and its corresponding

definition, reference and intensity rating. Panelists were encouraged to review their tray of references to

refresh themselves as to the intensity and rating of the references for each attribute. Panelists were allowed to

review their references for as long as they wanted. After reviewing all of the references, panelists moved to

the booths to individually rate the samples. Testing took place in individual booths, with positive airflow and

temperature set to 70°F. All samples were presented in black double old-fashion glasses (Threshold, Target)

labeled with random three digit codes and evaluated under red lighting. The samples in each set were

presented to the panelists in random order. The ballot was presented to panelists using Compusense seven as

the interface in the booth and Compusense five to launch the test.

When panelists entered the booth they were instructed to enter their panelist code, and then indicated to the

session facilitator that they were ready for the test by turning on their light. Panelists received both samples at

the same time and were told to check the sample codes before proceeding with the test. Panelists were cued

with instructions for each modality, and strictly followed the evaluation protocols they had developed for

each modality as outlined in day 15. Panelists were allowed to re-evaluate the sample provided that they

follow the rinse protocol between tastings. Panelists were given a 0-15 point categorical scale with which to

rate the samples, with 0 being no perception of the attribute and 15 being the greatest intensity of the

attribute in the rum samples. Panelists were provided with the attribute, definition, reference and reference

score for each term. After evaluating the first samples, panelists were given a 2 minute break where they were

instructed to follow the rinse procedure to cleanse their pallet before the next sample. After rating the

second sample panelists indicated to the facilitator that they were had concluded the test.

264

Panelists then came back a second time during the day, at least 30 minutes after finishing the first session to

evaluate a second sample set. Panelists were asked to review any references that they felt gave them trouble

but were told they did not have to review every attribute. The second sample was also evaluated in the

booths.


Panelists again attended two 30 minute sessions and the procedure was the same as outlined for day 19

(session 3: Ron Abuelo 7 year and El Dorado 12 Year, session 4: Bacardi White and Diplomatico Reserva 12

year).



(session 5: Bacardi Gold and Appleton Estate 12 Year, session 6: Diplomatico Reserva 12 year and Ron

Abuelo 7 year).



(session 7: Appleton Estate V/X and Dictador XO Insolent, session 8: El Dorado 12 Year and Appleton

Estate 12 year).

Day 23: Individual booth testing of rum sample set 9

Panelists attended only one 30 minute session and the procedure was the same as outlined for day 19 (session

9: Bacardi White and Ron Zacapa). Throughout the five days of booth testing, rums were assigned to the test

sessions randomly, making sure that all 9 rums were presented one time before the second set of evaluated

took place. All rums were evaluated in duplicate.

265

Appendix P: Detailed day-by-day procedural explanation of ethanol dilution descriptive analysis panel

Day 1: Introductory session and term generation for rum samples

After signing in, the facilitator gave a short presentation introducing the panelists to the purpose of the

second DA panel and a review of basic DA procedures and protocols. Panelists were informed that samples

would be presented in sets of three, with the same three samples always appearing together, samples sets

consisted of the same rum at different dillutions. The importance of following the rinse protocol was highly

stressed as differences between samples would be smaller than the first panel. The rinse protocol was the

same as the previous panel, bread, warm water and room temperature water, with the panelist expectorating

all of the rinses and samples. Panelists were then presented with three samples, all dilutions of Ron Abuelo 7

year (straight rum, 1:2 dilution with 40% ethanol, 1:2 dilution with water). Panelists were instructed to

evaluate all three samples and develop terms for the different modalities (aroma, aroma-by-mouth, taste,

mouthfeel, and aftertaste) along with corresponding definitions and references. They were presented with a

term generation worksheet (Appendix H) to aid in the development of terms. During discussion, panelists

mentioned all terms, definitions and references that had been generated. Through discussion, the panelists

compiled an extensive list of possible attributes and corresponding references to be investigated the following

day. The panel was then dismissed and the panel facilitator compiled the final list of terms, definitions and

references. The initial list consisted of 55 terms and 39 references. The panel facilitator then purchased the

references the panel had identified from local stores.

Day 2: Term generation and reference refinement

After signing in, the panelists were presented will all of the references they had requested during term

generation on day 24. Panelists determined if these references correctly represented the attribute detected in

the rums in terms of both quality and concentration. After examining all of the references, panelists were

presented with a different dilution series, this time the Diplomatico Reserva 12 year rum (straight rum, 1:2

dilution with 40% ethanol, 1:2 dilution with water), to continue term generation and reference refinement.

After panelists had a chance to evaluate all of the rum samples, panelists discussed their findings and put

forth new terms and references. Panelists reached consensus on terms and references that needed to be

added, modified or eliminated for the next session. Panelists were then dismissed and the facilitator compiled

the observations from the session in preparation for the next day.

Day 3: Term generation and reference refinement

After signing in, the panelists were presented will all of the references they had kept from the previous day

and the new references generated on day 25. Panelists determined if these references correctly represented the

attribute detected in the rums in terms of both quality and concentration. After examining all of the

266

references, panelists were presented with the Ron Abuelo 7 year dilution series to continue term generation

and reference refinement. After panelists had a chance to evaluate all of the rum samples, panelists discussed

their findings and put forth new terms and references. Panelists had stated that they were having a difficult

time generating at silky mouthfeel reference. The facilitator provided the panel with several potential

references to aid in reference selection, including 2% ,whole, almond and coconut milks. Panelists identified

almond milk as having the same silky mouthfeel as found in the rums. Panelists reached consensus on terms

and references that needed to be added, modified or eliminated for the next session. As a group, the panelists

and facilitator discussed which attributes were identified in each rum sample. Panelists were then dismissed

and the facilitator compiled the observations from the session in preparation for the next day.

Day 4: Reference refinement

After signing in, the panelists were presented will the references from the previous day, modified if necessary.

Panelists determined if these references correctly represented the attribute detected in the rums in terms of

both quality and concentration. After examining all of the references, panelists were presented with the

Diplomatico Reserva 12 year rum dilution series to continue term generation and reference refinement.

Panelists reached consensus on terms and references that needed to be added, modified or eliminated for the

next session. As a group, the panelists and facilitator discussed which attributes were identified in each rum

sample. Panelists were then dismissed and the facilitator compiled the observations from the session in

preparation for scaling the next day.

Day 5: Scaling of aroma, taste and mouthfeel attributes

After signing in, panelists were presented with the references they had identified during term generation and

were presented with the Diplomatico Reserva 12 year dilution set. Panelists were reminded of the reference

rating procedure, as presented on day 7. Paneilsts were asked to select the sample with the highest intensity

of that attribute and rate it a 15 and then rate the reference and other two samples accordingly. Panelists rated

the intensity of the aroma, mouthfeel and taste attributes. Panelists then discussed which references needed to

be modified for the next day. The facilitator then dismissed the panel and then compiled the reference and

samples score to provide feedback to the panelists the following day. Any reference that received a score of

15 was modified to be presented to the panelists again the next day.

Day 6: Scaling of aroma-by-mouth and aftertaste attributes

After signing in, panelists were presented with the references they had identified during term generation and

were presented with the Ron Abuelo year dilution set. Panelists were reminded of the reference rating

procedure, as presented on day 7. Panelists rated the intensity of the aroma-by-mouth, and aftertaste

attributes, as well as any attributes that had scored about a 15 the previous day. Panelists discussed which

267

references needed to be modified for the following day. The facilitator then dismissed the panel and then

compiled the reference and samples score to provide feedback to the panelists the following day. Any

reference that received a score of 15 was modified to be presented to the panelists again the next day.

Day 7: Scaling

After signing in, panelists were presented with the modified references as well as the Ron Abuelo 7 year

dilution set. Panelists were asked to review their references and then score the references that had been

modified from the previous day. Panelists then scored the three rums for each attribute in order as they

would in the booth, using their reference as an anchor point for the scale. Panelists then discussed terms they

were struggling with and worked towards rating those terms uniformly. The panel eliminated terms that were

difficult to identify or didn’t change between samples in either rum set. The facilitator then dismissed the

panel and then compiled the reference and samples score to provide feedback to the panelists the following

day.

Day 8: Scaling

After signing in, panelists were presented with the references and their scores from the precious day to aid in

panel training. Panelists were also presented with the Diplomatico Reserva 12 year dilution set. Panelists were

asked to rate the samples using the references as anchors on the scale. The facilitator and panel then

discussed terms the group was struggling with. Panelists decided to eliminate the grassy aroma and floral

aroma-by-mouth terms. The facilitator then dismissed the panel and compiled the sample scores to provide

feedback and aid in training the following day.

Day 9: Scaling


panel training. Panelists were also presented with the Ron Abuelo 7 year dilution set. Panelists were asked to

rate the samples using the references as anchors on the scale. The facilitator and panel then discussed terms

the group was struggling with, and panelists were free to adjust or remove any terms. The facilitator then

dismissed the panel and compiled the sample scores to provide feedback and aid in training the following day.

Day 10: Scaling


panel training. Panelists were again presented with the Ron Abuelo 7 year dilution set to aid in training and

allow panelists to evaluate their repeatability when scoring the samples. Panelists were asked to rate the

samples using the references as anchors on the scale. Once the panelists had finished rating the samples the

facilitator disclosed which samples corresponded to each other from the previous day. The facilitator and

268

panel then discussed terms the group was struggling with, and panelists were free to adjust or remove any

terms. The facilitator then dismissed the panel and compiled the sample scores to provide feedback and aid in

training the following day.

Day 11: Booth Practice

Panelists attended one booth practice session, to evaluate the Diplomatico Reserva dilution set. When

panelists arrived, the reviewed their references and then proceeded to the booth for testing. Booth testing

conditions were the same as for the first session with the panelists receiving three samples per set and only

rating one sample set for booth practice.

Day 12: Scaling Final Day

After signing in, panelists were presented with their scores from the previous day booth testing in order to aid

in training. Panelists were also presented with their references and the Diplomatico Reserva 12 year dilution

set, the same set they evaluated in booth testing in order to better work towards group consistency and work

on precision. Panelists rated the samples for the different modalities using the references as anchors for the

scale. After consolidating the rating for the day, panelists discussed the samples and compared their results to

their scores from the previous day. Panelists finalized the list of terms, definitions, references and reference

scores. The finalized list consisted of 24 total attributes: 8 aroma , 5 aroma-by-mouth, 2 taste, 4 mouthfeel

and 5 aftertaste attributes. The facilitator then dismissed the panel and compiled the data to prepare for booth

testing.

Day 13: Booth Testing

When panelists entered, they were presented with their finalized reference list. After reviewing the references,

panelists proceeded to the booth. Booth testing conditions were the same as in the first study, except this

time panelists were presented with a dilution sample set consisting of three rums, instead of the two rums like

before. Panelists evaluated both the Ron Abuelo and Diplomatico Reserva sample sets. Sample presentation

within the dilution set was randomized as well as which dilution set the panelists received first. Panelists

evaluated bo Panelists evaluated two sample sets, with a minimum of a thirty minute break between the end

of one evaluation session and the beginning of the next.

Day 14: Booth Testing

Panelists arrived for booth testing, which proceeded in the same fashion as presented above in day 36.

Day 15: Wrap-up and panelist payment

269

When panelists arrived, they were compensated for their participation in the study. Once all panelists had

been paid, the facilitator presented a short PowerPoint revealing to the panelists the brands of rums they had

been evaluating for the past 8 weeks. Panelists were allowed to ask any question they had about the samples

or DA panel procedures. Panelists were then dismissed.

270

Appendix Q: Permission statements

understanding the complex aroma chemistry of …

Documents